The best NLP papers of 2013
The best NLP papers of 2013 include "Distributed Representations of Words and Phrases and their Compositionality" and "Efficient Estimation of Word Representations in Vector Space", both focusing on improving the quality and speed of training word vectors. Other notable papers cover topics like speech recognition with deep recurrent neural networks, generating sequences with recurrent neural networks, and exploiting similarities among languages for machine translation.
1. Distributed Representations of Words and Phrases and their Compositionality
30 833 citations
The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several extensions that improve both the quality of the vectors and the training speed. By subsampling of the frequent words we obtain significant speedup and also learn more regular word representations. We also describe a simple alternative to the hierarchical softmax called negative sampling. An inherent limitation of word representations is their indifference to word order and their inability to represent idiomatic phrases. For example, the meanings of "Canada" and "Air" cannot be easily combined to obtain "Air Canada". Motivated by this example, we present a simple method for finding phrases in text, and show that learning good vector representations for millions of phrases is possible.
2. Efficient Estimation of Word Representations in Vector Space
27 844 citations
We propose two novel model architectures for computing continuous vector representations of words from very large data sets. The quality of these representations is measured in a word similarity task, and the results are compared to the previously best performing techniques based on different types of neural networks. We observe large improvements in accuracy at much lower computational cost, i.e. it takes less than a day to learn high quality word vectors from a 1.6 billion words data set. Furthermore, we show that these vectors provide state-of-the-art performance on our test set for measuring syntactic and semantic word similarities.
3. Speech Recognition with Deep Recurrent Neural Networks
7 974 citations
Recurrent neural networks (RNNs) are a powerful model for sequential data. End-to-end training methods such as Connectionist Temporal Classification make it possible to train RNNs for sequence labelling problems where the input-output alignment is unknown. The combination of these methods with the Long Short-term Memory RNN architecture has proved particularly fruitful, delivering state-of-the-art results in cursive handwriting recognition. However RNN performance in speech recognition has so far been disappointing, with better results returned by deep feedforward networks. This paper investigates \emph{deep recurrent neural networks}, which combine the multiple levels of representation that have proved so effective in deep networks with the flexible use of long range context that empowers RNNs. When trained end-to-end with suitable regularisation, we find that deep Long Short-term Memory RNNs achieve a test set error of 17.7% on the TIMIT phoneme recognition benchmark, which to our knowledge is the best recorded score.
4. Generating Sequences With Recurrent Neural Networks
3 704 citations
This paper shows how Long Short-term Memory recurrent neural networks can be used to generate complex sequences with long-range structure, simply by predicting one data point at a time. The approach is demonstrated for text (where the data are discrete) and online handwriting (where the data are real-valued). It is then extended to handwriting synthesis by allowing the network to condition its predictions on a text sequence. The resulting system is able to generate highly realistic cursive handwriting in a wide variety of styles.
5. Crowdsourcing a Word-Emotion Association Lexicon
1 975 citations
Even though considerable attention has been given to the polarity of words (positive and negative) and the creation of large polarity lexicons, research in emotion analysis has had to rely on limited and small emotion lexicons. In this paper we show how the combined strength and wisdom of the crowds can be used to generate a large, high-quality, word-emotion and word-polarity association lexicon quickly and inexpensively. We enumerate the challenges in emotion annotation in a crowdsourcing scenario and propose solutions to address them. Most notably, in addition to questions about emotions associated with terms, we show how the inclusion of a word choice question can discourage malicious data entry, help identify instances where the annotator may not be familiar with the target term (allowing us to reject such annotations), and help obtain annotations at sense level (rather than at word level). We conducted experiments on how to formulate the emotion-annotation questions, and show that asking if a term is associated with an emotion leads to markedly higher inter-annotator agreement than that obtained by asking if a term evokes an emotion.
6. Exploiting Similarities among Languages for Machine Translation
1 516 citations
Dictionaries and phrase tables are the basis of modern statistical machine translation systems. This paper develops a method that can automate the process of generating and extending dictionaries and phrase tables. Our method can translate missing word and phrase entries by learning language structures based on large monolingual data and mapping between languages from small bilingual data. It uses distributed representation of words and learns a linear mapping between vector spaces of languages. Despite its simplicity, our method is surprisingly effective: we can achieve almost 90% precision@5 for translation of words between English and Spanish. This method makes little assumption about the languages, so it can be used to extend and refine dictionaries and translation tables for any language pairs.
7. NRC-Canada: Building the State-of-the-Art in Sentiment Analysis of Tweets
1 054 citations
In this paper, we describe how we created two state-of-the-art SVM classifiers, one to detect the sentiment of messages such as tweets and SMS (message-level task) and one to detect the sentiment of a term within a submissions stood first in both tasks on tweets, obtaining an F-score of 69.02 in the message-level task and 88.93 in the term-level task. We implemented a variety of surface-form, semantic, and sentiment features. with sentiment-word hashtags, and one from tweets with emoticons. In the message-level task, the lexicon-based features provided a gain of 5 F-score points over all others. Both of our systems can be replicated us available resources.
8. One Billion Word Benchmark for Measuring Progress in Statistical Language Modeling
1 007 citations
We propose a new benchmark corpus to be used for measuring progress in statistical language modeling. With almost one billion words of training data, we hope this benchmark will be useful to quickly evaluate novel language modeling techniques, and to compare their contribution when combined with other advanced techniques. We show performance of several well-known types of language models, with the best results achieved with a recurrent neural network based language model. The baseline unpruned Kneser-Ney 5-gram model achieves perplexity 67.6; a combination of techniques leads to 35% reduction in perplexity, or 10% reduction in cross-entropy (bits), over that baseline. The benchmark is available as a code.google.com project; besides the scripts needed to rebuild the training/held-out data, it also makes available log-probability values for each word in each of ten held-out data sets, for each of the baseline n-gram models.
9. Linear models and linear mixed effects models in R with linguistic applications
549 citations
This text is a conceptual introduction to mixed effects modeling with linguistic applications, using the R programming environment. The reader is introduced to linear modeling and assumptions, as well as to mixed effects/multilevel modeling, including a discussion of random intercepts, random slopes and likelihood ratio tests. The example used throughout the text focuses on the phonetic analysis of voice pitch data.
10. Polyglot: Distributed Word Representations for Multilingual NLP
459 citations
Distributed word representations (word embeddings) have recently contributed to competitive performance in language modeling and several NLP tasks. In this work, we train word embeddings for more than 100 languages using their corresponding Wikipedias. We quantitatively demonstrate the utility of our word embeddings by using them as the sole features for training a part of speech tagger for a subset of these languages. We find their performance to be competitive with near state-of-art methods in English, Danish and Swedish. Moreover, we investigate the semantic features captured by these embeddings through the proximity of word groupings. We will release these embeddings publicly to help researchers in the development and enhancement of multilingual applications.
11. A Computational Approach to Politeness with Application to Social Factors
367 citations
We propose a computational framework for identifying linguistic aspects of politeness. Our starting point is a new corpus of requests annotated for politeness, which we use to evaluate aspects of politeness theory and to uncover new interactions between politeness markers and context. These findings guide our construction of a classifier with domain-independent lexical and syntactic features operationalizing key components of politeness theory, such as indirection, deference, impersonalization and modality. Our classifier achieves close to human performance and is effective across domains. We use our framework to study the relationship between politeness and social power, showing that polite Wikipedia editors are more likely to achieve high status through elections, but, once elevated, they become less polite. We see a similar negative correlation between politeness and power on Stack Exchange, where users at the top of the reputation scale are less polite than those at the bottom. Finally, we apply our classifier to a preliminary analysis of politeness variation by gender and community.
12. Sentiment Analysis in the News
305 citations
Recent years have brought a significant growth in the volume of research in sentiment analysis, mostly on highly subjective text types (movie or product reviews). The main difference these texts have with news articles is that their target is clearly defined and unique across the text. Following different annotation efforts and the analysis of the issues encountered, we realised that news opinion mining is different from that of other text types. We identified three subtasks that need to be addressed: definition of the target; separation of the good and bad news content from the good and bad sentiment expressed on the target; and analysis of clearly marked opinion that is expressed explicitly, not needing interpretation or the use of world knowledge. Furthermore, we distinguish three different possible views on newspaper articles - author, reader and text, which have to be addressed differently at the time of analysing sentiment. Given these definitions, we present work on mining opinions about entities in English language news, in which (a) we test the relative suitability of various sentiment dictionaries and (b) we attempt to separate positive or negative opinion from good or bad news. In the experiments described here, we tested whether or not subject domain-defining vocabulary should be ignored. Results showed that this idea is more appropriate in the context of news opinion mining and that the approaches taking this into consideration produce a better performance.
13. Fast and accurate sentiment classification using an enhanced Naive Bayes model
259 citations
We have explored different methods of improving the accuracy of a Naive Bayes classifier for sentiment analysis. We observed that a combination of methods like negation handling, word n-grams and feature selection by mutual information results in a significant improvement in accuracy. This implies that a highly accurate and fast sentiment classifier can be built using a simple Naive Bayes model that has linear training and testing time complexities. We achieved an accuracy of 88.80% on the popular IMDB movie reviews dataset.
14. Recurrent Convolutional Neural Networks for Discourse Compositionality
251 citations
The compositionality of meaning extends beyond the single sentence. Just as words combine to form the meaning of sentences, so do sentences combine to form the meaning of paragraphs, dialogues and general discourse. We introduce both a sentence model and a discourse model corresponding to the two levels of compositionality. The sentence model adopts convolution as the central operation for composing semantic vectors and is based on a novel hierarchical convolutional neural network. The discourse model extends the sentence model and is based on a recurrent neural network that is conditioned in a novel way both on the current sentence and on the current speaker. The discourse model is able to capture both the sequentiality of sentences and the interaction between different speakers. Without feature engineering or pretraining and with simple greedy decoding, the discourse model coupled to the sentence model obtains state of the art performance on a dialogue act classification experiment.
15. Feature Learning in Deep Neural Networks - Studies on Speech Recognition Tasks
248 citations
Recent studies have shown that deep neural networks (DNNs) perform significantly better than shallow networks and Gaussian mixture models (GMMs) on large vocabulary speech recognition tasks. In this paper, we argue that the improved accuracy achieved by the DNNs is the result of their ability to extract discriminative internal representations that are robust to the many sources of variability in speech signals. We show that these representations become increasingly insensitive to small perturbations in the input with increasing network depth, which leads to better speech recognition performance with deeper networks. We also show that DNNs cannot extrapolate to test samples that are substantially different from the training examples. If the training data are sufficiently representative, however, internal features learned by the DNN are relatively stable with respect to speaker differences, bandwidth differences, and environment distortion. This enables DNN-based recognizers to perform as well or better than state-of-the-art systems based on GMMs or shallow networks without the need for explicit model adaptation or feature normalization.
16. Word Emdeddings through Hellinger PCA
248 citations
Word embeddings resulting from neural language models have been shown to be successful for a large variety of NLP tasks. However, such architecture might be difficult to train and time-consuming. Instead, we propose to drastically simplify the word embeddings computation through a Hellinger PCA of the word co-occurence matrix. We compare those new word embeddings with some well-known embeddings on NER and movie review tasks and show that we can reach similar or even better performance. Although deep learning is not really necessary for generating good word embeddings, we show that it can provide an easy way to adapt embeddings to specific tasks.
17. Good Debt or Bad Debt: Detecting Semantic Orientations in Economic Texts
247 citations
The use of robo-readers to analyze news texts is an emerging technology trend in computational finance. In recent research, a substantial effort has been invested to develop sophisticated financial polarity-lexicons that can be used to investigate how financial sentiments relate to future company performance. However, based on experience from other fields, where sentiment analysis is commonly applied, it is well-known that the overall semantic orientation of a sentence may differ from the prior polarity of individual words. The objective of this article is to investigate how semantic orientations can be better detected in financial and economic news by accommodating the overall phrase-structure information and domain-specific use of language. Our three main contributions are: (1) establishment of a human-annotated finance phrase-bank, which can be used as benchmark for training and evaluating alternative models; (2) presentation of a technique to enhance financial lexicons with attributes that help to identify expected direction of events that affect overall sentiment; (3) development of a linearized phrase-structure model for detecting contextual semantic orientations in financial and economic news texts. The relevance of the newly added lexicon features and the benefit of using the proposed learning-algorithm are demonstrated in a comparative study against previously used general sentiment models as well as the popular word frequency models used in recent financial studies. The proposed framework is parsimonious and avoids the explosion in feature-space caused by the use of conventional n-gram features.
18. Connecting Language and Knowledge Bases with Embedding Models for Relation Extraction
236 citations
This paper proposes a novel approach for relation extraction from free text which is trained to jointly use information from the text and from existing knowledge. Our model is based on two scoring functions that operate by learning low-dimensional embeddings of words and of entities and relationships from a knowledge base. We empirically show on New York Times articles aligned with Freebase relations that our approach is able to efficiently use the extra information provided by a large subset of Freebase data (4M entities, 23k relationships) to improve over existing methods that rely on text features alone.
19. Improvements to deep convolutional neural networks for LVCSR
218 citations
Deep Convolutional Neural Networks (CNNs) are more powerful than Deep Neural Networks (DNN), as they are able to better reduce spectral variation in the input signal. This has also been confirmed experimentally, with CNNs showing improvements in word error rate (WER) between 4-12% relative compared to DNNs across a variety of LVCSR tasks. In this paper, we describe different methods to further improve CNN performance. First, we conduct a deep analysis comparing limited weight sharing and full weight sharing with state-of-the-art features. Second, we apply various pooling strategies that have shown improvements in computer vision to an LVCSR speech task. Third, we introduce a method to effectively incorporate speaker adaptation, namely fMLLR, into log-mel features. Fourth, we introduce an effective strategy to use dropout during Hessian-free sequence training. We find that with these improvements, particularly with fMLLR and dropout, we are able to achieve an additional 2-3% relative improvement in WER on a 50-hour Broadcast News task over our previous best CNN baseline. On a larger 400-hour BN task, we find an additional 4-5% relative improvement over our previous best CNN baseline.
20. Cross-Recurrence Quantification Analysis of Categorical and Continuous Time Series: an R package
207 citations
This paper describes the R package crqa to perform cross-recurrence quantification analysis of two time series of either a categorical or continuous nature. Streams of behavioral information, from eye movements to linguistic elements, unfold over time. When two people interact, such as in conversation, they often adapt to each other, leading these behavioral levels to exhibit recurrent states. In dialogue, for example, interlocutors adapt to each other by exchanging interactive cues: smiles, nods, gestures, choice of words, and so on. In order for us to capture closely the goings-on of dynamic interaction, and uncover the extent of coupling between two individuals, we need to quantify how much recurrence is taking place at these levels. Methods available in crqa would allow researchers in cognitive science to pose such questions as how much are two people recurrent at some level of analysis, what is the characteristic lag time for one person to maximally match another, or whether one person is leading another. First, we set the theoretical ground to understand the difference between 'correlation' and 'co-visitation' when comparing two time series, using an aggregative or cross-recurrence approach. Then, we describe more formally the principles of cross-recurrence, and show with the current package how to carry out analyses applying them. We end the paper by comparing computational efficiency, and results' consistency, of crqa R package, with the benchmark MATLAB toolbox crptoolbox. We show perfect comparability between the two libraries on both levels.
21. Estimating Phoneme Class Conditional Probabilities from Raw Speech Signal using Convolutional Neural Networks
199 citations
In hybrid hidden Markov model/artificial neural networks (HMM/ANN) automatic speech recognition (ASR) system, the phoneme class conditional probabilities are estimated by first extracting acoustic features from the speech signal based on prior knowledge such as, speech perception or/and speech production knowledge, and, then modeling the acoustic features with an ANN. Recent advances in machine learning techniques, more specifically in the field of image processing and text processing, have shown that such divide and conquer strategy (i.e., separating feature extraction and modeling steps) may not be necessary. Motivated from these studies, in the framework of convolutional neural networks (CNNs), this paper investigates a novel approach, where the input to the ANN is raw speech signal and the output is phoneme class conditional probability estimates. On TIMIT phoneme recognition task, we study different ANN architectures to show the benefit of CNNs and compare the proposed approach against conventional approach where, spectral-based feature MFCC is extracted and modeled by a multilayer perceptron. Our studies show that the proposed approach can yield comparable or better phoneme recognition performance when compared to the conventional approach. It indicates that CNNs can learn features relevant for phoneme classification automatically from the raw speech signal.
22. Integrating Document Clustering and Topic Modeling
193 citations
Document clustering and topic modeling are two closely related tasks which can mutually benefit each other. Topic modeling can project documents into a topic space which facilitates effective document clustering. Cluster labels discovered by document clustering can be incorporated into topic models to extract local topics specific to each cluster and global topics shared by all clusters. In this paper, we propose a multi-grain clustering topic model (MGCTM) which integrates document clustering and topic modeling into a unified framework and jointly performs the two tasks to achieve the overall best performance. Our model tightly couples two components: a mixture component used for discovering latent groups in document collection and a topic model component used for mining multi-grain topics including local topics specific to each cluster and global topics shared across clusters.We employ variational inference to approximate the posterior of hidden variables and learn model parameters. Experiments on two datasets demonstrate the effectiveness of our model.
23. Domain and Function: A Dual-Space Model of Semantic Relations and Compositions
191 citations
Given appropriate representations of the semantic relations between carpenter and wood and between mason and stone (for example, vectors in a vector space model), a suitable algorithm should be able to recognize that these relations are highly similar (carpenter is to wood as mason is to stone; the relations are analogous). Likewise, with representations of dog, house, and kennel, an algorithm should be able to recognize that the semantic composition of dog and house, dog house, is highly similar to kennel (dog house and kennel are synonymous). It seems that these two tasks, recognizing relations and compositions, are closely connected. However, up to now, the best models for relations are significantly different from the best models for compositions. In this paper, we introduce a dual-space model that unifies these two tasks. This model matches the performance of the best previous models for relations and compositions. The dual-space model consists of a space for measuring domain similarity and a space for measuring function similarity. Carpenter and wood share the same domain, the domain of carpentry. Mason and stone share the same domain, the domain of masonry. Carpenter and mason share the same function, the function of artisans. Wood and stone share the same function, the function of materials. In the composition dog house, kennel has some domain overlap with both dog and house (the domains of pets and buildings). The function of kennel is similar to the function of house (the function of shelters). By combining domain and function similarities in various ways, we can model relations, compositions, and other aspects of semantics.
24. From Once Upon a Time to Happily Ever After: Tracking Emotions in Novels and Fairy Tales
185 citations
Today we have access to unprecedented amounts of literary texts. However, search still relies heavily on key words. In this paper, we show how sentiment analysis can be used in tandem with effective visualizations to quantify and track emotions in both individual books and across very large collections. We introduce the concept of emotion word density, and using the Brothers Grimm fairy tales as example, we show how collections of text can be organized for better search. Using the Google Books Corpus we show how to determine an entity's emotion associations from co-occurring words. Finally, we compare emotion words in fairy tales and novels, to show that fairy tales have a much wider range of emotion word densities than novels.
25. Tracking Sentiment in Mail: How Genders Differ on Emotional Axes
175 citations
With the widespread use of email, we now have access to unprecedented amounts of text that we ourselves have written. In this paper, we show how sentiment analysis can be used in tandem with effective visualizations to quantify and track emotions in many types of mail. We create a large word--emotion association lexicon by crowdsourcing, and use it to compare emotions in love letters, hate mail, and suicide notes. We show that there are marked differences across genders in how they use emotion words in work-place email. For example, women use many words from the joy--sadness axis, whereas men prefer terms from the fear--trust axis. Finally, we show visualizations that can help people track emotions in their emails.
26. Multilingual Distributed Representations without Word Alignment
153 citations
Distributed representations of meaning are a natural way to encode covariance relationships between words and phrases in NLP. By overcoming data sparsity problems, as well as providing information about semantic relatedness which is not available in discrete representations, distributed representations have proven useful in many NLP tasks. Recent work has shown how compositional semantic representations can successfully be applied to a number of monolingual applications such as sentiment analysis. At the same time, there has been some initial success in work on learning shared word-level representations across languages. We combine these two approaches by proposing a method for learning distributed representations in a multilingual setup. Our model learns to assign similar embeddings to aligned sentences and dissimilar ones to sentence which are not aligned while not requiring word alignments. We show that our representations are semantically informative and apply them to a cross-lingual document classification task where we outperform the previous state of the art. Further, by employing parallel corpora of multiple language pairs we find that our model learns representations that capture semantic relationships across languages for which no parallel data was used.
27. Multi-Step Regression Learning for Compositional Distributional Semantics
142 citations
We present a model for compositional distributional semantics related to the framework of Coecke et al. (2010), and emulating formal semantics by representing functions as tensors and arguments as vectors. We introduce a new learning method for tensors, generalising the approach of Baroni and Zamparelli (2010). We evaluate it on two benchmark data sets, and find it to outperform existing leading methods. We argue in our analysis that the nature of this learning method also renders it suitable for solving more subtle problems compositional distributional models might face.
28. DGT-TM: A freely Available Translation Memory in 22 Languages
142 citations
The European Commission's (EC) Directorate General for Translation, together with the EC's Joint Research Centre, is making available a large translation memory (TM; i.e. sentences and their professionally produced translations) covering twenty-two official European Union (EU) languages and their 231 language pairs. Such a resource is typically used by translation professionals in combination with TM software to improve speed and consistency of their translations. However, this resource has also many uses for translation studies and for language technology applications, including Statistical Machine Translation (SMT), terminology extraction, Named Entity Recognition (NER), multilingual classification and clustering, and many more. In this reference paper for DGT-TM, we introduce this new resource, provide statistics regarding its size, and explain how it was produced and how to use it.
29. Techniques for Feature Extraction In Speech Recognition System : A Comparative Study
126 citations
The time domain waveform of a speech signal carries all of the auditory information. From the phonological point of view, it little can be said on the basis of the waveform itself. However, past research in mathematics, acoustics, and speech technology have provided many methods for converting data that can be considered as information if interpreted correctly. In order to find some statistically relevant information from incoming data, it is important to have mechanisms for reducing the information of each segment in the audio signal into a relatively small number of parameters, or features. These features should describe each segment in such a characteristic way that other similar segments can be grouped together by comparing their features. There are enormous interesting and exceptional ways to describe the speech signal in terms of parameters. Though, they all have their strengths and weaknesses, we have presented some of the most used methods with their importance.
30. A Survey of Data Mining Techniques for Social Media Analysis
118 citations
Social network has gained remarkable attention in the last decade. Accessing social network sites such as Twitter, Facebook LinkedIn and Google+ through the internet and the web 2.0 technologies has become more affordable. People are becoming more interested in and relying on social network for information, news and opinion of other users on diverse subject matters. The heavy reliance on social network sites causes them to generate massive data characterised by three computational issues namely; size, noise and dynamism. These issues often make social network data very complex to analyse manually, resulting in the pertinent use of computational means of analysing them. Data mining provides a wide range of techniques for detecting useful knowledge from massive datasets like trends, patterns and rules [44]. Data mining techniques are used for information retrieval, statistical modelling and machine learning. These techniques employ data pre-processing, data analysis, and data interpretation processes in the course of data analysis. This survey discusses different data mining techniques used in mining diverse aspects of the social network over decades going from the historical techniques to the up-to-date models, including our novel technique named TRCM. All the techniques covered in this survey are listed in the Table.1 including the tools employed as well as names of their authors.
31. Description and Evaluation of Semantic Similarity Measures Approaches
115 citations
In recent years, semantic similarity measure has a great interest in Semantic Web and Natural Language Processing (NLP). Several similarity measures have been developed, being given the existence of a structured knowledge representation offered by ontologies and corpus which enable semantic interpretation of terms. Semantic similarity measures compute the similarity between concepts/terms included in knowledge sources in order to perform estimations. This paper discusses the existing semantic similarity methods based on structure, information content and feature approaches. Additionally, we present a critical evaluation of several categories of semantic similarity approaches based on two standard benchmarks. The aim of this paper is to give an efficient evaluation of all these measures which help researcher and practitioners to select the measure that best fit for their requirements.
32. Probabilistic Frame Induction
113 citations
In natural-language discourse, related events tend to appear near each other to describe a larger scenario. Such structures can be formalized by the notion of a frame (a.k.a. template), which comprises a set of related events and prototypical participants and event transitions. Identifying frames is a prerequisite for information extraction and natural language generation, and is usually done manually. Methods for inducing frames have been proposed recently, but they typically use ad hoc procedures and are difficult to diagnose or extend. In this paper, we propose the first probabilistic approach to frame induction, which incorporates frames, events, participants as latent topics and learns those frame and event transitions that best explain the text. The number of frames is inferred by a novel application of a split-merge method from syntactic parsing. In end-to-end evaluations from text to induced frames and extracted facts, our method produced state-of-the-art results while substantially reducing engineering effort.
33. Joint Video and Text Parsing for Understanding Events and Answering Queries
113 citations
We propose a framework for parsing video and text jointly for understanding events and answering user queries. Our framework produces a parse graph that represents the compositional structures of spatial information (objects and scenes), temporal information (actions and events) and causal information (causalities between events and fluents) in the video and text. The knowledge representation of our framework is based on a spatial-temporal-causal And-Or graph (S/T/C-AOG), which jointly models possible hierarchical compositions of objects, scenes and events as well as their interactions and mutual contexts, and specifies the prior probabilistic distribution of the parse graphs. We present a probabilistic generative model for joint parsing that captures the relations between the input video/text, their corresponding parse graphs and the joint parse graph. Based on the probabilistic model, we propose a joint parsing system consisting of three modules: video parsing, text parsing and joint inference. Video parsing and text parsing produce two parse graphs from the input video and text respectively. The joint inference module produces a joint parse graph by performing matching, deduction and revision on the video and text parse graphs. The proposed framework has the following objectives: Firstly, we aim at deep semantic parsing of video and text that goes beyond the traditional bag-of-words approaches; Secondly, we perform parsing and reasoning across the spatial, temporal and causal dimensions based on the joint S/T/C-AOG representation; Thirdly, we show that deep joint parsing facilitates subsequent applications such as generating narrative text descriptions and answering queries in the forms of who, what, when, where and why. We empirically evaluated our system based on comparison against ground-truth as well as accuracy of query answering and obtained satisfactory results.
34. Computing Lexical Contrast
105 citations
Knowing the degree of semantic contrast between words has widespread application in natural language processing, including machine translation, information retrieval, and dialogue systems. Manually-created lexicons focus on opposites, such as {\rm hot} and {\rm cold}. Opposites are of many kinds such as antipodals, complementaries, and gradable. However, existing lexicons often do not classify opposites into the different kinds. They also do not explicitly list word pairs that are not opposites but yet have some degree of contrast in meaning, such as {\rm warm} and {\rm cold} or {\rm tropical} and {\rm freezing}. We propose an automatic method to identify contrasting word pairs that is based on the hypothesis that if a pair of words, $A$ and $B$, are contrasting, then there is a pair of opposites, $C$ and $D$, such that $A$ and $C$ are strongly related and $B$ and $D$ are strongly related. (For example, there exists the pair of opposites {\rm hot} and {\rm cold} such that {\rm tropical} is related to {\rm hot,} and {\rm freezing} is related to {\rm cold}.) We will call this the contrast hypothesis. We begin with a large crowdsourcing experiment to determine the amount of human agreement on the concept of oppositeness and its different kinds. In the process, we flesh out key features of different kinds of opposites. We then present an automatic and empirical measure of lexical contrast that relies on the contrast hypothesis, corpus statistics, and the structure of a {\it Roget}-like thesaurus. We show that the proposed measure of lexical contrast obtains high precision and large coverage, outperforming existing methods.
35. Sentiment Analysis: How to Derive Prior Polarities from SentiWordNet
102 citations
Assigning a positive or negative score to a word out of context (i.e. a word's prior polarity) is a challenging task for sentiment analysis. In the literature, various approaches based on SentiWordNet have been proposed. In this paper, we compare the most often used techniques together with newly proposed ones and incorporate all of them in a learning framework to see whether blending them can further improve the estimation of prior polarity scores. Using two different versions of SentiWordNet and testing regression and classification models across tasks and datasets, our learning approach consistently outperforms the single metrics, providing a new state-of-the-art approach in computing words' prior polarity for sentiment analysis. We conclude our investigation showing interesting biases in calculated prior polarity scores when word Part of Speech and annotator gender are considered.
36. The most controversial topics in Wikipedia: A multilingual and geographical analysis
100 citations
We present, visualize and analyse the similarities and differences between the controversial topics related to "edit wars" identified in 10 different language versions of Wikipedia. After a brief review of the related work we describe the methods developed to locate, measure, and categorize the controversial topics in the different languages. Visualizations of the degree of overlap between the top 100 lists of most controversial articles in different languages and the content related to geographical locations will be presented. We discuss what the presented analysis and visualizations can tell us about the multicultural aspects of Wikipedia and practices of peer-production. Our results indicate that Wikipedia is more than just an encyclopaedia; it is also a window into convergent and divergent social-spatial priorities, interests and preferences.
37. Opinion Mining and Analysis: A survey
99 citations
The current research is focusing on the area of Opinion Mining also called as sentiment analysis due to sheer volume of opinion rich web resources such as discussion forums, review sites and blogs are available in digital form. One important problem in sentiment analysis of product reviews is to produce summary of opinions based on product features. We have surveyed and analyzed in this paper, various techniques that have been developed for the key tasks of opinion mining. We have provided an overall picture of what is involved in developing a software system for opinion mining on the basis of our survey and analysis.
38. Arabizi Detection and Conversion to Arabic
98 citations
Arabizi is Arabic text that is written using Latin characters. Arabizi is used to present both Modern Standard Arabic (MSA) or Arabic dialects. It is commonly used in informal settings such as social networking sites and is often with mixed with English. In this paper we address the problems of: identifying Arabizi in text and converting it to Arabic characters. We used word and sequence-level features to identify Arabizi that is mixed with English. We achieved an identification accuracy of 98.5%. As for conversion, we used transliteration mining with language modeling to generate equivalent Arabic text. We achieved 88.7% conversion accuracy, with roughly a third of errors being spelling and morphological variants of the forms in ground truth.
39. Towards a Formal Distributional Semantics: Simulating Logical Calculi with Tensors
97 citations
The development of compositional distributional models of semantics reconciling the empirical aspects of distributional semantics with the compositional aspects of formal semantics is a popular topic in the contemporary literature. This paper seeks to bring this reconciliation one step further by showing how the mathematical constructs commonly used in compositional distributional models, such as tensors and matrices, can be used to simulate different aspects of predicate logic. This paper discusses how the canonical isomorphism between tensors and multilinear maps can be exploited to simulate a full-blown quantifier-free predicate calculus using tensors. It provides tensor interpretations of the set of logical connectives required to model propositional calculi. It suggests a variant of these tensor calculi capable of modelling quantifiers, using few non-linear operations. It finally discusses the relation between these variants, and how this relation should constitute the subject of future work.
40. An introduction to the Europe Media Monitor family of applications
97 citations
Most large organizations have dedicated departments that monitor the media to keep up-to-date with relevant developments and to keep an eye on how they are represented in the news. Part of this media monitoring work can be automated. In the European Union with its 23 official languages, it is particularly important to cover media reports in many languages in order to capture the complementary news content published in the different countries. It is also important to be able to access the news content across languages and to merge the extracted information. We present here the four publicly accessible systems of the Europe Media Monitor (EMM) family of applications, which cover between 19 and 50 languages (see http://press.jrc.it/overview.html). We give an overview of their functionality and discuss some of the implications of the fact that they cover quite so many languages. We discuss design issues necessary to be able to achieve this high multilinguality, as well as the benefits of this multilinguality.
41. WebSets: Extracting Sets of Entities from the Web Using Unsupervised Information Extraction
96 citations
We describe a open-domain information extraction method for extracting concept-instance pairs from an HTML corpus. Most earlier approaches to this problem rely on combining clusters of distributionally similar terms and concept-instance pairs obtained with Hearst patterns. In contrast, our method relies on a novel approach for clustering terms found in HTML tables, and then assigning concept names to these clusters using Hearst patterns. The method can be efficiently applied to a large corpus, and experimental results on several datasets show that our method can accurately extract large numbers of concept-instance pairs.
42. Performance Investigation of Feature Selection Methods
94 citations
Sentiment analysis or opinion mining has become an open research domain after proliferation of Internet and Web 2.0 social media. People express their attitudes and opinions on social media including blogs, discussion forums, tweets, etc. and, sentiment analysis concerns about detecting and extracting sentiment or opinion from online text. Sentiment based text classification is different from topical text classification since it involves discrimination based on expressed opinion on a topic. Feature selection is significant for sentiment analysis as the opinionated text may have high dimensions, which can adversely affect the performance of sentiment analysis classifier. This paper explores applicability of feature selection methods for sentiment analysis and investigates their performance for classification in term of recall, precision and accuracy. Five feature selection methods (Document Frequency, Information Gain, Gain Ratio, Chi Squared, and Relief-F) and three popular sentiment feature lexicons (HM, GI and Opinion Lexicon) are investigated on movie reviews corpus with a size of 2000 documents. The experimental results show that Information Gain gave consistent results and Gain Ratio performs overall best for sentimental feature selection while sentiment lexicons gave poor performance. Furthermore, we found that performance of the classifier depends on appropriate number of representative feature selected from text.
43. The Expressive Power of Word Embeddings
93 citations
We seek to better understand the difference in quality of the several publicly released embeddings. We propose several tasks that help to distinguish the characteristics of different embeddings. Our evaluation of sentiment polarity and synonym/antonym relations shows that embeddings are able to capture surprisingly nuanced semantics even in the absence of sentence structure. Moreover, benchmarking the embeddings shows great variance in quality and characteristics of the semantics captured by the tested embeddings. Finally, we show the impact of varying the number of dimensions and the resolution of each dimension on the effective useful features captured by the embedding space. Our contributions highlight the importance of embeddings for NLP tasks and the effect of their quality on the final results.
44. Lambek vs. Lambek: Functorial Vector Space Semantics and String Diagrams for Lambek Calculus
92 citations
The Distributional Compositional Categorical (DisCoCat) model is a mathematical framework that provides compositional semantics for meanings of natural language sentences. It consists of a computational procedure for constructing meanings of sentences, given their grammatical structure in terms of compositional type-logic, and given the empirically derived meanings of their words. For the particular case that the meaning of words is modelled within a distributional vector space model, its experimental predictions, derived from real large scale data, have outperformed other empirically validated methods that could build vectors for a full sentence. This success can be attributed to a conceptually motivated mathematical underpinning, by integrating qualitative compositional type-logic and quantitative modelling of meaning within a category-theoretic mathematical framework. The type-logic used in the DisCoCat model is Lambek's pregroup grammar. Pregroup types form a posetal compact closed category, which can be passed, in a functorial manner, on to the compact closed structure of vector spaces, linear maps and tensor product. The diagrammatic versions of the equational reasoning in compact closed categories can be interpreted as the flow of word meanings within sentences. Pregroups simplify Lambek's previous type-logic, the Lambek calculus, which has been extensively used to formalise and reason about various linguistic phenomena. The apparent reliance of the DisCoCat on pregroups has been seen as a shortcoming. This paper addresses this concern, by pointing out that one may as well realise a functorial passage from the original type-logic of Lambek, a monoidal bi-closed category, to vector spaces, or to any other model of meaning organised within a monoidal bi-closed category. The corresponding string diagram calculus, due to Baez and Stay, now depicts the flow of word meanings.
45. Distributional semantics beyond words: Supervised learning of analogy and paraphrase
89 citations
There have been several efforts to extend distributional semantics beyond individual words, to measure the similarity of word pairs, phrases, and sentences (briefly, tuples; ordered sets of words, contiguous or noncontiguous). One way to extend beyond words is to compare two tuples using a function that combines pairwise similarities between the component words in the tuples. A strength of this approach is that it works with both relational similarity (analogy) and compositional similarity (paraphrase). However, past work required hand-coding the combination function for different tasks. The main contribution of this paper is that combination functions are generated by supervised learning. We achieve state-of-the-art results in measuring relational similarity between word pairs (SAT analogies and SemEval~2012 Task 2) and measuring compositional similarity between noun-modifier phrases and unigrams (multiple-choice paraphrase questions).
46. Quantum Entanglement in Concept Combinations
86 citations
Research in the application of quantum structures to cognitive science confirms that these structures quite systematically appear in the dynamics of concepts and their combinations and quantum-based models faithfully represent experimental data of situations where classical approaches are problematical. In this paper, we analyze the data we collected in an experiment on a specific conceptual combination, showing that Bell's inequalities are violated in the experiment. We present a new refined entanglement scheme to model these data within standard quantum theory rules, where 'entangled measurements and entangled evolutions' occur, in addition to the expected 'entangled states', and present a full quantum representation in complex Hilbert space of the data. This stronger form of entanglement in measurements and evolutions might have relevant applications in the foundations of quantum theory, as well as in the interpretation of nonlocality tests. It could indeed explain some non-negligible 'anomalies' identified in EPR-Bell experiments.
47. Semantic Measures for the Comparison of Units of Language, Concepts or Instances from Text and Knowledge Base Analysis
86 citations
Semantic measures are widely used today to estimate the strength of the semantic relationship between elements of various types: units of language (e.g., words, sentences, documents), concepts or even instances semantically characterized (e.g., diseases, genes, geographical locations). Semantic measures play an important role to compare such elements according to semantic proxies: texts and knowledge representations, which support their meaning or describe their nature. Semantic measures are therefore essential for designing intelligent agents which will for example take advantage of semantic analysis to mimic human ability to compare abstract or concrete objects. This paper proposes a comprehensive survey of the broad notion of semantic measure for the comparison of units of language, concepts or instances based on semantic proxy analyses. Semantic measures generalize the well-known notions of semantic similarity, semantic relatedness and semantic distance, which have been extensively studied by various communities over the last decades (e.g., Cognitive Sciences, Linguistics, and Artificial Intelligence to mention a few).
48. A probabilistic framework for analysing the compositionality of conceptual combinations
78 citations
Conceptual combination performs a fundamental role in creating the broad range of compound phrases utilized in everyday language. This article provides a novel probabilistic framework for assessing whether the semantics of conceptual combinations are compositional, and so can be considered as a function of the semantics of the constituent concepts, or not. While the systematicity and productivity of language provide a strong argument in favor of assuming compositionality, this very assumption is still regularly questioned in both cognitive science and philosophy. Additionally, the principle of semantic compositionality is underspecified, which means that notions of both "strong" and "weak" compositionality appear in the literature. Rather than adjudicating between different grades of compositionality, the framework presented here contributes formal methods for determining a clear dividing line between compositional and non-compositional semantics. In addition, we suggest that the distinction between these is contextually sensitive. Utilizing formal frameworks developed for analyzing composite systems in quantum theory, we present two methods that allow the semantics of conceptual combinations to be classified as "compositional" or "non-compositional". Compositionality is first formalised by factorising the joint probability distribution modeling the combination, where the terms in the factorisation correspond to individual concepts. This leads to the necessary and sufficient condition for the joint probability distribution to exist. A failure to meet this condition implies that the underlying concepts cannot be modeled in a single probability space when considering their combination, and the combination is thus deemed "non-compositional". The formal analysis methods are demonstrated by applying them to an empirical study of twenty-four non-lexicalised conceptual combinations.
49. Automatic Labeling for Entity Extraction in Cyber Security
78 citations
Timely analysis of cyber-security information necessitates automated information extraction from unstructured text. While state-of-the-art extraction methods produce extremely accurate results, they require ample training data, which is generally unavailable for specialized applications, such as detecting security related entities; moreover, manual annotation of corpora is very costly and often not a viable solution. In response, we develop a very precise method to automatically label text from several data sources by leveraging related, domain-specific, structured data and provide public access to a corpus annotated with cyber-security entities. Next, we implement a Maximum Entropy Model trained with the average perceptron on a portion of our corpus ($\sim$750,000 words) and achieve near perfect precision, recall, and accuracy, with training times under 17 seconds.
50. Inferring Informational Goals from Free-Text Queries: A Bayesian Approach
76 citations
People using consumer software applications typically do not use technical jargon when querying an online database of help topics. Rather, they attempt to communicate their goals with common words and phrases that describe software functionality in terms of structure and objects they understand. We describe a Bayesian approach to modeling the relationship between words in a user's query for assistance and the informational goals of the user. After reviewing the general method, we describe several extensions that center on integrating additional distinctions and structure about language usage and user goals into the Bayesian models.
51. Compression as a universal principle of animal behavior
76 citations
A key aim in biology and psychology is to identify fundamental principles underpinning the behavior of animals, including humans. Analyses of human language and the behavior of a range of non-human animal species have provided evidence for a common pattern underlying diverse behavioral phenomena: words follow Zipf's law of brevity (the tendency of more frequently used words to be shorter), and conformity to this general pattern has been seen in the behavior of a number of other animals. It has been argued that the presence of this law is a sign of efficient coding in the information theoretic sense. However, no strong direct connection has been demonstrated between the law and compression, the information theoretic principle of minimizing the expected length of a code. Here we show that minimizing the expected code length implies that the length of a word cannot increase as its frequency increases. Furthermore, we show that the mean code length or duration is significantly small in human language, and also in the behavior of other species in all cases where agreement with the law of brevity has been found. We argue that compression is a general principle of animal behavior, that reflects selection for efficiency of coding.
52. Multilinguals and Wikipedia Editing
76 citations
This article analyzes one month of edits to Wikipedia in order to examine the role of users editing multiple language editions (referred to as multilingual users). Such multilingual users may serve an important function in diffusing information across different language editions of the encyclopedia, and prior work has suggested this could reduce the level of self-focus bias in each edition. This study finds multilingual users are much more active than their single-edition (monolingual) counterparts. They are found in all language editions, but smaller-sized editions with fewer users have a higher percentage of multilingual users than larger-sized editions. About a quarter of multilingual users always edit the same articles in multiple languages, while just over 40% of multilingual users edit different articles in different languages. When non-English users do edit a second language edition, that edition is most frequently English. Nonetheless, several regional and linguistic cross-editing patterns are also present.
53. Sentiment in New York City: A High Resolution Spatial and Temporal View
71 citations
Measuring public sentiment is a key task for researchers and policymakers alike. The explosion of available social media data allows for a more time-sensitive and geographically specific analysis than ever before. In this paper we analyze data from the micro-blogging site Twitter and generate a sentiment map of New York City. We develop a classifier specifically tuned for 140-character Twitter messages, or tweets, using key words, phrases and emoticons to determine the mood of each tweet. This method, combined with geotagging provided by users, enables us to gauge public sentiment on extremely fine-grained spatial and temporal scales. We find that public mood is generally highest in public parks and lowest at transportation hubs, and locate other areas of strong sentiment such as cemeteries, medical centers, a jail, and a sewage facility. Sentiment progressively improves with proximity to Times Square. Periodic patterns of sentiment fluctuate on both a daily and a weekly scale: more positive tweets are posted on weekends than on weekdays, with a daily peak in sentiment around midnight and a nadir between 9:00 a.m. and noon.
54. Supervised Topical Key Phrase Extraction of News Stories using Crowdsourcing, Light Filtering and Co-reference Normalization
68 citations
Fast and effective automated indexing is critical for search and personalized services. Key phrases that consist of one or more words and represent the main concepts of the document are often used for the purpose of indexing. In this paper, we investigate the use of additional semantic features and pre-processing steps to improve automatic key phrase extraction. These features include the use of signal words and freebase categories. Some of these features lead to significant improvements in the accuracy of the results. We also experimented with 2 forms of document pre-processing that we call light filtering and co-reference normalization. Light filtering removes sentences from the document, which are judged peripheral to its main content. Co-reference normalization unifies several written forms of the same named entity into a unique form. We also needed a "Gold Standard" - a set of labeled documents for training and evaluation. While the subjective nature of key phrase selection precludes a true "Gold Standard", we used Amazon's Mechanical Turk service to obtain a useful approximation. Our data indicates that the biggest improvements in performance were due to shallow semantic features, news categories, and rhetorical signals (nDCG 78.47% vs. 68.93%). The inclusion of deeper semantic features such as Freebase sub-categories was not beneficial by itself, but in combination with pre-processing, did cause slight improvements in the nDCG scores.
55. JRC-Names: A freely available, highly multilingual named entity resource
68 citations
This paper describes a new, freely available, highly multilingual named entity resource for person and organisation names that has been compiled over seven years of large-scale multilingual news analysis combined with Wikipedia mining, resulting in 205,000 per-son and organisation names plus about the same number of spelling variants written in over 20 different scripts and in many more languages. This resource, produced as part of the Europe Media Monitor activity (EMM, http://emm.newsbrief.eu/overview.html), can be used for a number of purposes. These include improving name search in databases or on the internet, seeding machine learning systems to learn named entity recognition rules, improve machine translation results, and more. We describe here how this resource was created; we give statistics on its current size; we address the issue of morphological inflection; and we give details regarding its functionality. Updates to this resource will be made available daily.
56. Learning New Facts From Knowledge Bases With Neural Tensor Networks and Semantic Word Vectors
64 citations
Knowledge bases provide applications with the benefit of easily accessible, systematic relational knowledge but often suffer in practice from their incompleteness and lack of knowledge of new entities and relations. Much work has focused on building or extending them by finding patterns in large unannotated text corpora. In contrast, here we mainly aim to complete a knowledge base by predicting additional true relationships between entities, based on generalizations that can be discerned in the given knowledgebase. We introduce a neural tensor network (NTN) model which predicts new relationship entries that can be added to the database. This model can be improved by initializing entity representations with word vectors learned in an unsupervised fashion from text, and when doing this, existing relations can even be queried for entities that were not present in the database. Our model generalizes and outperforms existing models for this problem, and can classify unseen relationships in WordNet with an accuracy of 75.8%.
57. Early Stage Influenza Detection from Twitter
63 citations
Influenza is an acute respiratory illness that occurs virtually every year and results in substantial disease, death and expense. Detection of Influenza in its earliest stage would facilitate timely action that could reduce the spread of the illness. Existing systems such as CDC and EISS which try to collect diagnosis data, are almost entirely manual, resulting in about two-week delays for clinical data acquisition. Twitter, a popular microblogging service, provides us with a perfect source for early-stage flu detection due to its real- time nature. For example, when a flu breaks out, people that get the flu may post related tweets which enables the detection of the flu breakout promptly. In this paper, we investigate the real-time flu detection problem on Twitter data by proposing Flu Markov Network (Flu-MN): a spatio-temporal unsupervised Bayesian algorithm based on a 4 phase Markov Network, trying to identify the flu breakout at the earliest stage. We test our model on real Twitter datasets from the United States along with baselines in multiple applications, such as real-time flu breakout detection, future epidemic phase prediction, or Influenza-like illness (ILI) physician visits. Experimental results show the robustness and effectiveness of our approach. We build up a real time flu reporting system based on the proposed approach, and we are hopeful that it would help government or health organizations in identifying flu outbreaks and facilitating timely actions to decrease unnecessary mortality.
58. Implementation Of Back-Propagation Neural Network For Isolated Bangla Speech Recognition
60 citations
This paper is concerned with the development of Back-propagation Neural Network for Bangla Speech Recognition. In this paper, ten bangla digits were recorded from ten speakers and have been recognized. The features of these speech digits were extracted by the method of Mel Frequency Cepstral Coefficient (MFCC) analysis. The mfcc features of five speakers were used to train the network with Back propagation algorithm. The mfcc features of ten bangla digit speeches, from 0 to 9, of another five speakers were used to test the system. All the methods and algorithms used in this research were implemented using the features of Turbo C and C++ languages. From our investigation it is seen that the developed system can successfully encode and analyze the mfcc features of the speech signal to recognition. The developed system achieved recognition rate about 96.332% for known speakers (i.e., speaker dependent) and 92% for unknown speakers (i.e., speaker independent).
59. Probing the statistical properties of unknown texts: application to the Voynich Manuscript
59 citations
While the use of statistical physics methods to analyze large corpora has been useful to unveil many patterns in texts, no comprehensive investigation has been performed investigating the properties of statistical measurements across different languages and texts. In this study we propose a framework that aims at determining if a text is compatible with a natural language and which languages are closest to it, without any knowledge of the meaning of the words. The approach is based on three types of statistical measurements, i.e. obtained from first-order statistics of word properties in a text, from the topology of complex networks representing text, and from intermittency concepts where text is treated as a time series. Comparative experiments were performed with the New Testament in 15 different languages and with distinct books in English and Portuguese in order to quantify the dependency of the different measurements on the language and on the story being told in the book. The metrics found to be informative in distinguishing real texts from their shuffled versions include assortativity, degree and selectivity of words. As an illustration, we analyze an undeciphered medieval manuscript known as the Voynich Manuscript. We show that it is mostly compatible with natural languages and incompatible with random texts. We also obtain candidates for key-words of the Voynich Manuscript which could be helpful in the effort of deciphering it. Because we were able to identify statistical measurements that are more dependent on the syntax than on the semantics, the framework may also serve for text analysis in language-dependent applications.
60. A Comparison of Named Entity Recognition Tools Applied to Biographical Texts
59 citations
Named entity recognition (NER) is a popular domain of natural language processing. For this reason, many tools exist to perform this task. Amongst other points, they differ in the processing method they rely upon, the entity types they can detect, the nature of the text they can handle, and their input/output formats. This makes it difficult for a user to select an appropriate NER tool for a specific situation. In this article, we try to answer this question in the context of biographic texts. For this matter, we first constitute a new corpus by annotating Wikipedia articles. We then select publicly available, well known and free for research NER tools for comparison: Stanford NER, Illinois NET, OpenCalais NER WS and Alias-i LingPipe. We apply them to our corpus, assess their performances and compare them. When considering overall performances, a clear hierarchy emerges: Stanford has the best results, followed by LingPipe, Illionois and OpenCalais. However, a more detailed evaluation performed relatively to entity types and article categories highlights the fact their performances are diversely influenced by those factors. This complementarity opens an interesting perspective regarding the combination of these individual tools in order to improve performance.
61. Structure-semantics interplay in complex networks and its effects on the predictability of similarity in texts
56 citations
There are different ways to define similarity for grouping similar texts into clusters, as the concept of similarity may depend on the purpose of the task. For instance, in topic extraction similar texts mean those within the same semantic field, whereas in author recognition stylistic features should be considered. In this study, we introduce ways to classify texts employing concepts of complex networks, which may be able to capture syntactic, semantic and even pragmatic features. The interplay between the various metrics of the complex networks is analyzed with three applications, namely identification of machine translation (MT) systems, evaluation of quality of machine translated texts and authorship recognition. We shall show that topological features of the networks representing texts can enhance the ability to identify MT systems in particular cases. For evaluating the quality of MT texts, on the other hand, high correlation was obtained with methods capable of capturing the semantics. This was expected because the golden standards used are themselves based on word co-occurrence. Notwithstanding, the Katz similarity, which involves semantic and structure in the comparison of texts, achieved the highest correlation with the NIST measurement, indicating that in some cases the combination of both approaches can improve the ability to quantify quality in MT. In authorship recognition, again the topological features were relevant in some contexts, though for the books and authors analyzed good results were obtained with semantic features as well. Because hybrid approaches encompassing semantic and topological features have not been extensively used, we believe that the methodology proposed here may be useful to enhance text classification considerably, as it combines well-established strategies.
62. Arabic text summarization based on latent semantic analysis to enhance arabic documents clustering
54 citations
Arabic Documents Clustering is an important task for obtaining good results with the traditional Information Retrieval (IR) systems especially with the rapid growth of the number of online documents present in Arabic language. Documents clustering aim to automatically group similar documents in one cluster using different similarity/distance measures. This task is often affected by the documents length, useful information on the documents is often accompanied by a large amount of noise, and therefore it is necessary to eliminate this noise while keeping useful information to boost the performance of Documents clustering. In this paper, we propose to evaluate the impact of text summarization using the Latent Semantic Analysis Model on Arabic Documents Clustering in order to solve problems cited above, using five similarity/distance measures: Euclidean Distance, Cosine Similarity, Jaccard Coefficient, Pearson Correlation Coefficient and Averaged Kullback-Leibler Divergence, for two times: without and with stemming. Our experimental results indicate that our proposed approach effectively solves the problems of noisy information and documents length, and thus significantly improve the clustering performance.
63. The risks of mixing dependency lengths from sequences of different length
51 citations
Mixing dependency lengths from sequences of different length is a common practice in language research. However, the empirical distribution of dependency lengths of sentences of the same length differs from that of sentences of varying length and the distribution of dependency lengths depends on sentence length for real sentences and also under the null hypothesis that dependencies connect vertices located in random positions of the sequence. This suggests that certain results, such as the distribution of syntactic dependency lengths mixing dependencies from sentences of varying length, could be a mere consequence of that mixing. Furthermore, differences in the global averages of dependency length (mixing lengths from sentences of varying length) for two different languages do not simply imply a priori that one language optimizes dependency lengths better than the other because those differences could be due to differences in the distribution of sentence lengths and other factors.
64. LDC Arabic Treebanks and Associated Corpora: Data Divisions Manual
50 citations
The Linguistic Data Consortium (LDC) has developed hundreds of data corpora for natural language processing (NLP) research. Among these are a number of annotated treebank corpora for Arabic. Typically, these corpora consist of a single collection of annotated documents. NLP research, however, usually requires multiple data sets for the purposes of training models, developing techniques, and final evaluation. Therefore it becomes necessary to divide the corpora used into the required data sets (divisions). This document details a set of rules that have been defined to enable consistent divisions for old and new Arabic treebanks (ATB) and related corpora.
65. The placement of the head that minimizes online memory: a complex systems approach
47 citations
It is well known that the length of a syntactic dependency determines its online memory cost. Thus, the problem of the placement of a head and its dependents (complements or modifiers) that minimizes online memory is equivalent to the problem of the minimum linear arrangement of a star tree. However, how that length is translated into cognitive cost is not known. This study shows that the online memory cost is minimized when the head is placed at the center, regardless of the function that transforms length into cost, provided only that this function is strictly monotonically increasing. Online memory defines a quasi-convex adaptive landscape with a single central minimum if the number of elements is odd and two central minima if that number is even. We discuss various aspects of the dynamics of word order of subject (S), verb (V) and object (O) from a complex systems perspective and suggest that word orders tend to evolve by swapping adjacent constituents from an initial or early SOV configuration that is attracted towards a central word order by online memory minimization. We also suggest that the stability of SVO is due to at least two factors, the quasi-convex shape of the adaptive landscape in the online memory dimension and online memory adaptations that avoid regression to SOV. Although OVS is also optimal for placing the verb at the center, its low frequency is explained by its long distance to the seminal SOV in the permutation space.
66. Architecture of an Ontology-Based Domain-Specific Natural Language Question Answering System
46 citations
Question answering (QA) system aims at retrieving precise information from a large collection of documents against a query. This paper describes the architecture of a Natural Language Question Answering (NLQA) system for a specific domain based on the ontological information, a step towards semantic web question answering. The proposed architecture defines four basic modules suitable for enhancing current QA capabilities with the ability of processing complex questions. The first module was the question processing, which analyses and classifies the question and also reformulates the user query. The second module allows the process of retrieving the relevant documents. The next module processes the retrieved documents, and the last module performs the extraction and generation of a response. Natural language processing techniques are used for processing the question and documents and also for answer extraction. Ontology and domain knowledge are used for reformulating queries and identifying the relations. The aim of the system is to generate short and specific answer to the question that is asked in the natural language in a specific domain. We have achieved 94 % accuracy of natural language question answering in our implementation.
67. Automatic case acquisition from texts for process-oriented case-based reasoning
45 citations
This paper introduces a method for the automatic acquisition of a rich case representation from free text for process-oriented case-based reasoning. Case engineering is among the most complicated and costly tasks in implementing a case-based reasoning system. This is especially so for process-oriented case-based reasoning, where more expressive case representations are generally used and, in our opinion, actually required for satisfactory case adaptation. In this context, the ability to acquire cases automatically from procedural texts is a major step forward in order to reason on processes. We therefore detail a methodology that makes case acquisition from processes described as free text possible, with special attention given to assembly instruction texts. This methodology extends the techniques we used to extract actions from cooking recipes. We argue that techniques taken from natural language processing are required for this task, and that they give satisfactory results. An evaluation based on our implemented prototype extracting workflows from recipe texts is provided.
68. End-to-end Phoneme Sequence Recognition using Convolutional Neural Networks
44 citations
Most phoneme recognition state-of-the-art systems rely on a classical neural network classifiers, fed with highly tuned features, such as MFCC or PLP features. Recent advances in ``deep learning'' approaches questioned such systems, but while some attempts were made with simpler features such as spectrograms, state-of-the-art systems still rely on MFCCs. This might be viewed as a kind of failure from deep learning approaches, which are often claimed to have the ability to train with raw signals, alleviating the need of hand-crafted features. In this paper, we investigate a convolutional neural network approach for raw speech signals. While convolutional architectures got tremendous success in computer vision or text processing, they seem to have been let down in the past recent years in the speech processing field. We show that it is possible to learn an end-to-end phoneme sequence classifier system directly from raw signal, with similar performance on the TIMIT and WSJ datasets than existing systems based on MFCC, questioning the need of complex hand-crafted features on large datasets.
69. Sentiment Analysis : A Literature Survey
43 citations
Our day-to-day life has always been influenced by what people think. Ideas and opinions of others have always affected our own opinions. The explosion of Web 2.0 has led to increased activity in Podcasting, Blogging, Tagging, Contributing to RSS, Social Bookmarking, and Social Networking. As a result there has been an eruption of interest in people to mine these vast resources of data for opinions. Sentiment Analysis or Opinion Mining is the computational treatment of opinions, sentiments and subjectivity of text. In this report, we take a look at the various challenges and applications of Sentiment Analysis. We will discuss in details various approaches to perform a computational treatment of sentiments and opinions. Various supervised or data-driven techniques to SA like Na\"ive Byes, Maximum Entropy, SVM, and Voted Perceptrons will be discussed and their strengths and drawbacks will be touched upon. We will also see a new dimension of analyzing sentiments by Cognitive Psychology mainly through the work of Janyce Wiebe, where we will see ways to detect subjectivity, perspective in narrative and understanding the discourse structure. We will also study some specific topics in Sentiment Analysis and the contemporary works in those areas.
70. "Not not bad" is not "bad": A distributional account of negation
43 citations
With the increasing empirical success of distributional models of compositional semantics, it is timely to consider the types of textual logic that such models are capable of capturing. In this paper, we address shortcomings in the ability of current models to capture logical operations such as negation. As a solution we propose a tripartite formulation for a continuous vector space representation of semantics and subsequently use this representation to develop a formal compositional notion of negation within such models.
71. Colourful Language: Measuring Word-Colour Associations
43 citations
Since many real-world concepts are associated with colour, for example danger with red, linguistic information is often complimented with the use of appropriate colours in information visualization and product marketing. Yet, there is no comprehensive resource that captures concept-colour associations. We present a method to create a large word-colour association lexicon by crowdsourcing. We focus especially on abstract concepts and emotions to show that even though they cannot be physically visualized, they too tend to have strong colour associations. Finally, we show how word-colour associations manifest themselves in language, and quantify usefulness of co-occurrence and polarity cues in automatically detecting colour associations.
72. Can recursive neural tensor networks learn logical reasoning?
43 citations
Recursive neural network models and their accompanying vector representations for words have seen success in an array of increasingly semantically sophisticated tasks, but almost nothing is known about their ability to accurately capture the aspects of linguistic meaning that are necessary for interpretation or reasoning. To evaluate this, I train a recursive model on a new corpus of constructed examples of logical reasoning in short sentences, like the inference of "some animal walks" from "some dog walks" or "some cat walks," given that dogs and cats are animals. This model learns representations that generalize well to new types of reasoning pattern in all but a few cases, a result which is promising for the ability of learned representation models to capture logical reasoning.
73. Word sense disambiguation via high order of learning in complex networks
42 citations
Complex networks have been employed to model many real systems and as a modeling tool in a myriad of applications. In this paper, we use the framework of complex networks to the problem of supervised classification in the word disambiguation task, which consists in deriving a function from the supervised (or labeled) training data of ambiguous words. Traditional supervised data classification takes into account only topological or physical features of the input data. On the other hand, the human (animal) brain performs both low- and high-level orders of learning and it has facility to identify patterns according to the semantic meaning of the input data. In this paper, we apply a hybrid technique which encompasses both types of learning in the field of word sense disambiguation and show that the high-level order of learning can really improve the accuracy rate of the model. This evidence serves to demonstrate that the internal structures formed by the words do present patterns that, generally, cannot be correctly unveiled by only traditional techniques. Finally, we exhibit the behavior of the model for different weights of the low- and high-level classifiers by plotting decision boundaries. This study helps one to better understand the effectiveness of the model.
74. A Multilingual Semantic Wiki Based on Attempto Controlled English and Grammatical Framework
42 citations
We describe a semantic wiki system with an underlying controlled natural language grammar implemented in Grammatical Framework (GF). The grammar restricts the wiki content to a well-defined subset of Attempto Controlled English (ACE), and facilitates a precise bidirectional automatic translation between ACE and language fragments of a number of other natural languages, making the wiki content accessible multilingually. Additionally, our approach allows for automatic translation into the Web Ontology Language (OWL), which enables automatic reasoning over the wiki content. The developed wiki environment thus allows users to build, query and view OWL knowledge bases via a user-friendly multilingual natural language interface. As a further feature, the underlying multilingual grammar is integrated into the wiki and can be collaboratively edited to extend the vocabulary of the wiki or even customize its sentence structures. This work demonstrates the combination of the existing technologies of Attempto Controlled English and Grammatical Framework, and is implemented as an extension of the existing semantic wiki engine AceWiki.
75. JRC EuroVoc Indexer JEX - A freely available multi-label categorisation tool
42 citations
EuroVoc (2012) is a highly multilingual thesaurus consisting of over 6,700 hierarchically organised subject domains used by European Institutions and many authorities in Member States of the European Union (EU) for the classification and retrieval of official documents. JEX is JRC-developed multi-label classification software that learns from manually labelled data to automatically assign EuroVoc descriptors to new documents in a profile-based category-ranking task. The JEX release consists of trained classifiers for 22 official EU languages, of parallel training data in the same languages, of an interface that allows viewing and amending the assignment results, and of a module that allows users to re-train the tool on their own document collections. JEX allows advanced users to change the document representation so as to possibly improve the categorisation result through linguistic pre-processing. JEX can be used as a tool for interactive EuroVoc descriptor assignment to increase speed and consistency of the human categorisation process, or it can be used fully automatically. The output of JEX is a language-independent EuroVoc feature vector lending itself also as input to various other Language Technology tasks, including cross-lingual clustering and classification, cross-lingual plagiarism detection, sentence selection and ranking, and more.
76. The Royal Birth of 2013: Analysing and Visualising Public Sentiment in the UK Using Twitter
39 citations
Analysis of information retrieved from microblogging services such as Twitter can provide valuable insight into public sentiment in a geographic region. This insight can be enriched by visualising information in its geographic context. Two underlying approaches for sentiment analysis are dictionary-based and machine learning. The former is popular for public sentiment analysis, and the latter has found limited use for aggregating public sentiment from Twitter data. The research presented in this paper aims to extend the machine learning approach for aggregating public sentiment. To this end, a framework for analysing and visualising public sentiment from a Twitter corpus is developed. A dictionary-based approach and a machine learning approach are implemented within the framework and compared using one UK case study, namely the royal birth of 2013. The case study validates the feasibility of the framework for analysis and rapid visualisation. One observation is that there is good correlation between the results produced by the popular dictionary-based approach and the machine learning approach when large volumes of tweets are analysed. However, for rapid analysis to be possible faster methods need to be developed using big data techniques and parallel methods.
77. Hubiness, length, crossings and their relationships in dependency trees
38 citations
Here tree dependency structures are studied from three different perspectives: their degree variance (hubiness), the mean dependency length and the number of dependency crossings. Bounds that reveal pairwise dependencies among these three metrics are derived. Hubiness (the variance of degrees) plays a central role: the mean dependency length is bounded below by hubiness while the number of crossings is bounded above by hubiness. Our findings suggest that the online memory cost of a sentence might be determined not just by the ordering of words but also by the hubiness of the underlying structure. The 2nd moment of degree plays a crucial role that is reminiscent of its role in large complex networks.
78. Rule Based Transliteration Scheme for English to Punjabi
38 citations
Machine Transliteration has come out to be an emerging and a very important research area in the field of machine translation. Transliteration basically aims to preserve the phonological structure of words. Proper transliteration of name entities plays a very significant role in improving the quality of machine translation. In this paper we are doing machine transliteration for English-Punjabi language pair using rule based approach. We have constructed some rules for syllabification. Syllabification is the process to extract or separate the syllable from the words. In this we are calculating the probabilities for name entities (Proper names and location). For those words which do not come under the category of name entities, separate probabilities are being calculated by using relative frequency through a statistical machine translation toolkit known as MOSES. Using these probabilities we are transliterating our input text from English to Punjabi.
79. Development of Marathi Part of Speech Tagger Using Statistical Approach
38 citations
Part-of-speech (POS) tagging is a process of assigning the words in a text corresponding to a particular part of speech. A fundamental version of POS tagging is the identification of words as nouns, verbs, adjectives etc. For processing natural languages, Part of Speech tagging is a prominent tool. It is one of the simplest as well as most constant and statistical model for many NLP applications. POS Tagging is an initial stage of linguistics, text analysis like information retrieval, machine translator, text to speech synthesis, information extraction etc. In POS Tagging we assign a Part of Speech tag to each word in a sentence and literature. Various approaches have been proposed to implement POS taggers. In this paper we present a Marathi part of speech tagger. It is morphologically rich language. Marathi is spoken by the native people of Maharashtra. The general approach used for development of tagger is statistical using Unigram, Bigram, Trigram and HMM Methods. It presents a clear idea about all the algorithms with suitable examples. It also introduces a tag set for Marathi which can be used for tagging Marathi text. In this paper we have shown the development of the tagger as well as compared to check the accuracy of taggers output. The three Marathi POS taggers viz. Unigram, Bigram, Trigram and HMM gives the accuracy of 77.38%, 90.30%, 91.46% and 93.82% respectively.
80. Seeing What You're Told: Sentence-Guided Activity Recognition In Video
37 citations
We present a system that demonstrates how the compositional structure of events, in concert with the compositional structure of language, can interplay with the underlying focusing mechanisms in video action recognition, thereby providing a medium, not only for top-down and bottom-up integration, but also for multi-modal integration between vision and language. We show how the roles played by participants (nouns), their characteristics (adjectives), the actions performed (verbs), the manner of such actions (adverbs), and changing spatial relations between participants (prepositions) in the form of whole sentential descriptions mediated by a grammar, guides the activity-recognition process. Further, the utility and expressiveness of our framework is demonstrated by performing three separate tasks in the domain of multi-activity videos: sentence-guided focus of attention, generation of sentential descriptions of video, and query-based video search, simply by leveraging the framework in different manners.
81. Unveiling the relationship between complex networks metrics and word senses
36 citations
The automatic disambiguation of word senses (i.e., the identification of which of the meanings is used in a given context for a word that has multiple meanings) is essential for such applications as machine translation and information retrieval, and represents a key step for developing the so-called Semantic Web. Humans disambiguate words in a straightforward fashion, but this does not apply to computers. In this paper we address the problem of Word Sense Disambiguation (WSD) by treating texts as complex networks, and show that word senses can be distinguished upon characterizing the local structure around ambiguous words. Our goal was not to obtain the best possible disambiguation system, but we nevertheless found that in half of the cases our approach outperforms traditional shallow methods. We show that the hierarchical connectivity and clustering of words are usually the most relevant features for WSD. The results reported here shine light on the relationship between semantic and structural parameters of complex networks. They also indicate that when combined with traditional techniques the complex network approach may be useful to enhance the discrimination of senses in large texts
82. Zero-Shot Learning for Semantic Utterance Classification
36 citations
We propose a novel zero-shot learning method for semantic utterance classification (SUC). It learns a classifier $f: X \to Y$ for problems where none of the semantic categories $Y$ are present in the training set. The framework uncovers the link between categories and utterances using a semantic space. We show that this semantic space can be learned by deep neural networks trained on large amounts of search engine query log data. More precisely, we propose a novel method that can learn discriminative semantic features without supervision. It uses the zero-shot learning framework to guide the learning of the semantic features. We demonstrate the effectiveness of the zero-shot semantic learning algorithm on the SUC dataset collected by (Tur, 2012). Furthermore, we achieve state-of-the-art results by combining the semantic features with a supervised method.
83. Complex networks analysis of language complexity
35 citations
Methods from statistical physics, such as those involving complex networks, have been increasingly used in quantitative analysis of linguistic phenomena. In this paper, we represented pieces of text with different levels of simplification in co-occurrence networks and found that topological regularity correlated negatively with textual complexity. Furthermore, in less complex texts the distance between concepts, represented as nodes, tended to decrease. The complex networks metrics were treated with multivariate pattern recognition techniques, which allowed us to distinguish between original texts and their simplified versions. For each original text, two simplified versions were generated manually with increasing number of simplification operations. As expected, distinction was easier for the strongly simplified versions, where the most relevant metrics were node strength, shortest paths and diversity. Also, the discrimination of complex texts was improved with higher hierarchical network metrics, thus pointing to the usefulness of considering wider contexts around the concepts. Though the accuracy rate in the distinction was not as high as in methods using deep linguistic knowledge, the complex network approach is still useful for a rapid screening of texts whenever assessing complexity is essential to guarantee accessibility to readers with limited reading ability
84. ManTIME: Temporal expression identification and normalization in the TempEval-3 challenge
35 citations
This paper describes a temporal expression identification and normalization system, ManTIME, developed for the TempEval-3 challenge. The identification phase combines the use of conditional random fields along with a post-processing identification pipeline, whereas the normalization phase is carried out using NorMA, an open-source rule-based temporal normalizer. We investigate the performance variation with respect to different feature types. Specifically, we show that the use of WordNet-based features in the identification task negatively affects the overall performance, and that there is no statistically significant difference in using gazetteers, shallow parsing and propositional noun phrases labels on top of the morphological features. On the test data, the best run achieved 0.95 (P), 0.85 (R) and 0.90 (F1) in the identification phase. Normalization accuracies are 0.84 (type attribute) and 0.77 (value attribute). Surprisingly, the use of the silver data (alone or in addition to the gold annotated ones) does not improve the performance.
85. PACE: Pattern Accurate Computationally Efficient Bootstrapping for Timely Discovery of Cyber-Security Concepts
35 citations
Public disclosure of important security information, such as knowledge of vulnerabilities or exploits, often occurs in blogs, tweets, mailing lists, and other online sources months before proper classification into structured databases. In order to facilitate timely discovery of such knowledge, we propose a novel semi-supervised learning algorithm, PACE, for identifying and classifying relevant entities in text sources. The main contribution of this paper is an enhancement of the traditional bootstrapping method for entity extraction by employing a time-memory trade-off that simultaneously circumvents a costly corpus search while strengthening pattern nomination, which should increase accuracy. An implementation in the cyber-security domain is discussed as well as challenges to Natural Language Processing imposed by the security domain.
86. Sockpuppet Detection in Wikipedia: A Corpus of Real-World Deceptive Writing for Linking Identities
35 citations
This paper describes the corpus of sockpuppet cases we gathered from Wikipedia. A sockpuppet is an online user account created with a fake identity for the purpose of covering abusive behavior and/or subverting the editing regulation process. We used a semi-automated method for crawling and curating a dataset of real sockpuppet investigation cases. To the best of our knowledge, this is the first corpus available on real-world deceptive writing. We describe the process for crawling the data and some preliminary results that can be used as baseline for benchmarking research. The dataset will be released under a Creative Commons license from our project website: http://docsig.cis.uab.edu.
87. Text segmentation with character-level text embeddings
34 citations
Learning word representations has recently seen much success in computational linguistics. However, assuming sequences of word tokens as input to linguistic analysis is often unjustified. For many languages word segmentation is a non-trivial task and naturally occurring text is sometimes a mixture of natural language strings and other character data. We propose to learn text representations directly from raw character sequences by training a Simple recurrent Network to predict the next character in text. The network uses its hidden layer to evolve abstract representations of the character sequences it sees. To demonstrate the usefulness of the learned text embeddings, we use them as features in a supervised character level text segmentation and labeling task: recognizing spans of text containing programming language code. By using the embeddings as features we are able to substantially improve over a baseline which uses only surface character n-grams.
88. Development of Yes/No Arabic Question Answering System
32 citations
Developing Question Answering systems has been one of the important research issues because it requires insights from a variety of disciplines,including,Artificial Intelligence,Information Retrieval, Information Extraction,Natural Language Processing, and Psychology.In this paper we realize a formal model for a lightweight semantic based open domain yes/no Arabic question answering system based on paragraph retrieval with variable length. We propose a constrained semantic representation. Using an explicit unification framework based on semantic similarities and query expansion synonyms and antonyms.This frequently improves the precision of the system. Employing the passage retrieval system achieves a better precision by retrieving more paragraphs that contain relevant answers to the question; It significantly reduces the amount of text to be processed by the system.
89. Using Nuances of Emotion to Identify Personality
32 citations
Past work on personality detection has shown that frequency of lexical categories such as first person pronouns, past tense verbs, and sentiment words have significant correlations with personality traits. In this paper, for the first time, we show that fine affect (emotion) categories such as that of excitement, guilt, yearning, and admiration are significant indicators of personality. Additionally, we perform experiments to show that the gains provided by the fine affect categories are not obtained by using coarse affect categories alone or with specificity features alone. We employ these features in five SVM classifiers for detecting five personality traits through essays. We find that the use of fine emotion features leads to statistically significant improvement over a competitive baseline, whereas the use of coarse affect and specificity features does not.
90. Category-Theoretic Quantitative Compositional Distributional Models of Natural Language Semantics
32 citations
This thesis is about the problem of compositionality in distributional semantics. Distributional semantics presupposes that the meanings of words are a function of their occurrences in textual contexts. It models words as distributions over these contexts and represents them as vectors in high dimensional spaces. The problem of compositionality for such models concerns itself with how to produce representations for larger units of text by composing the representations of smaller units of text. This thesis focuses on a particular approach to this compositionality problem, namely using the categorical framework developed by Coecke, Sadrzadeh, and Clark, which combines syntactic analysis formalisms with distributional semantic representations of meaning to produce syntactically motivated composition operations. This thesis shows how this approach can be theoretically extended and practically implemented to produce concrete compositional distributional models of natural language semantics. It furthermore demonstrates that such models can perform on par with, or better than, other competing approaches in the field of natural language processing. There are three principal contributions to computational linguistics in this thesis. The first is to extend the DisCoCat framework on the syntactic front and semantic front, incorporating a number of syntactic analysis formalisms and providing learning procedures allowing for the generation of concrete compositional distributional models. The second contribution is to evaluate the models developed from the procedures presented here, showing that they outperform other compositional distributional models present in the literature. The third contribution is to show how using category theory to solve linguistic problems forms a sound basis for research, illustrated by examples of work on this topic, that also suggest directions for future research.
91. Learning Semantic Representations for the Phrase Translation Model
32 citations
This paper presents a novel semantic-based phrase translation model. A pair of source and target phrases are projected into continuous-valued vector representations in a low-dimensional latent semantic space, where their translation score is computed by the distance between the pair in this new space. The projection is performed by a multi-layer neural network whose weights are learned on parallel training data. The learning is aimed to directly optimize the quality of end-to-end machine translation results. Experimental evaluation has been performed on two Europarl translation tasks, English-French and German-English. The results show that the new semantic-based phrase translation model significantly improves the performance of a state-of-the-art phrase-based statistical machine translation sys-tem, leading to a gain of 0.7-1.0 BLEU points.
92. I Wish I Didn't Say That! Analyzing and Predicting Deleted Messages in Twitter
31 citations
Twitter has become a major source of data for social media researchers. One important aspect of Twitter not previously considered are {\em deletions} -- removal of tweets from the stream. Deletions can be due to a multitude of reasons such as privacy concerns, rashness or attempts to undo public statements. We show how deletions can be automatically predicted ahead of time and analyse which tweets are likely to be deleted and how.
93. NILE: Fast Natural Language Processing for Electronic Health Records
31 citations
Objective: Narrative text in Electronic health records (EHR) contain rich information for medical and data science studies. This paper introduces the design and performance of Narrative Information Linear Extraction (NILE), a natural language processing (NLP) package for EHR analysis that we share with the medical informatics community. Methods: NILE uses a modified prefix-tree search algorithm for named entity recognition, which can detect prefix and suffix sharing. The semantic analyses are implemented as rule-based finite state machines. Analyses include negation, location, modification, family history, and ignoring. Result: The processing speed of NILE is hundreds to thousands times faster than existing NLP software for medical text. The accuracy of presence analysis of NILE is on par with the best performing models on the 2010 i2b2/VA NLP challenge data. Conclusion: The speed, accuracy, and being able to operate via API make NILE a valuable addition to the NLP software for medical informatics and data science.
94. A Hybrid Approach to Extract Keyphrases from Medical Documents
30 citations
Keyphrases are the phrases, consisting of one or more words, representing the important concepts in the articles. Keyphrases are useful for a variety of tasks such as text summarization, automatic indexing, clustering/classification, text mining etc. This paper presents a hybrid approach to keyphrase extraction from medical documents. The keyphrase extraction approach presented in this paper is an amalgamation of two methods: the first one assigns weights to candidate keyphrases based on an effective combination of features such as position, term frequency, inverse document frequency and the second one assign weights to candidate keyphrases using some knowledge about their similarities to the structure and characteristics of keyphrases available in the memory (stored list of keyphrases). An efficient candidate keyphrase identification method as the first component of the proposed keyphrase extraction system has also been introduced in this paper. The experimental results show that the proposed hybrid approach performs better than some state-of-the art keyphrase extraction approaches.
95. Bidirectional Recursive Neural Networks for Token-Level Labeling with Structure
30 citations
Recently, deep architectures, such as recurrent and recursive neural networks have been successfully applied to various natural language processing tasks. Inspired by bidirectional recurrent neural networks which use representations that summarize the past and future around an instance, we propose a novel architecture that aims to capture the structural information around an input, and use it to label instances. We apply our method to the task of opinion expression extraction, where we employ the binary parse tree of a sentence as the structure, and word vector representations as the initial representation of a single token. We conduct preliminary experiments to investigate its performance and compare it to the sequential approach.
96. A quantum teleportation inspired algorithm produces sentence meaning from word meaning and grammatical structure
29 citations
We discuss an algorithm which produces the meaning of a sentence given meanings of its words, and its resemblance to quantum teleportation. In fact, this protocol was the main source of inspiration for this algorithm which has many applications in the area of Natural Language Processing.
97. Even the Abstract have Colour: Consensus in Word-Colour Associations
29 citations
Colour is a key component in the successful dissemination of information. Since many real-world concepts are associated with colour, for example danger with red, linguistic information is often complemented with the use of appropriate colours in information visualization and product marketing. Yet, there is no comprehensive resource that captures concept-colour associations. We present a method to create a large word-colour association lexicon by crowdsourcing. A word-choice question was used to obtain sense-level annotations and to ensure data quality. We focus especially on abstract concepts and emotions to show that even they tend to have strong colour associations. Thus, using the right colours can not only improve semantic coherence, but also inspire the desired emotional response.
98. Cornell SPF: Cornell Semantic Parsing Framework
29 citations
The Cornell Semantic Parsing Framework (SPF) is a learning and inference framework for mapping natural language to formal representation of its meaning.
99. Language learning from positive evidence, reconsidered: A simplicity-based approach
28 citations
Children learn their native language by exposure to their linguistic and communicative environment, but apparently without requiring that their mistakes are corrected. Such learning from positive evidence has been viewed as raising logical problems for language acquisition. In particular, without correction, how is the child to recover from conjecturing an over-general grammar, which will be consistent with any sentence that the child hears? There have been many proposals concerning how this logical problem can be dissolved. Here, we review recent formal results showing that the learner has sufficient data to learn successfully from positive evidence, if it favours the simplest encoding of the linguistic input. Results include the ability to learn a linguistic prediction, grammaticality judgements, language production, and form-meaning mappings. The simplicity approach can also be scaled-down to analyse the ability to learn a specific linguistic constructions, and is amenable to empirical test as a framework for describing human language acquisition.
100. An open diachronic corpus of historical Spanish: annotation criteria and automatic modernisation of spelling
28 citations
The IMPACT-es diachronic corpus of historical Spanish compiles over one hundred books --containing approximately 8 million words-- in addition to a complementary lexicon which links more than 10 thousand lemmas with attestations of the different variants found in the documents. This textual corpus and the accompanying lexicon have been released under an open license (Creative Commons by-nc-sa) in order to permit their intensive exploitation in linguistic research. Approximately 7% of the words in the corpus (a selection aimed at enhancing the coverage of the most frequent word forms) have been annotated with their lemma, part of speech, and modern equivalent. This paper describes the annotation criteria followed and the standards, based on the Text Encoding Initiative recommendations, used to the represent the texts in digital form. As an illustration of the possible synergies between diachronic textual resources and linguistic research, we describe the application of statistical machine translation techniques to infer probabilistic context-sensitive rules for the automatic modernisation of spelling. The automatic modernisation with this type of statistical methods leads to very low character error rates when the output is compared with the supervised modern version of the text.