Distributed Word Representations Research Articles

A comprehensive analysis of static word embeddings for Turkish

Word embeddings are fixed-length, dense and distributed word representations that are used in natural language processing (NLP) applications. There are basically two types of word embedding models which are non-contextual (static) models and contextual models. The former method generates a single embedding for a word regardless of its context, while the latter method produces distinct embeddings for a word based on the specific contexts in which it appears. There are plenty of works that compare contextual and non-contextual embedding models within their respective groups in different languages. However, the number of studies that compare the models in these two groups with each other is very few and there is no such study in Turkish. This process necessitates converting contextual embeddings into static embeddings. In this paper, we compare and evaluate the performance of several contextual and non-contextual models in both intrinsic and extrinsic evaluation settings for Turkish. We make a fine-grained comparison by analyzing the syntactic and semantic capabilities of the models separately. The results of the analyses provide insights about the suitability of different embedding models in different types of NLP tasks. We also build a Turkish word embedding repository comprising the embedding models used in this work, which may serve as a valuable resource for researchers and practitioners in the field of Turkish NLP. We make the word embeddings, scripts, and evaluation datasets publicly available.

Exploring Latent Themes-Analysis of various Topic Modelling Algorithms

This research explores the effectiveness of four common topic modelling methods for identifying latent themes and topics in unstructured text data: Latent Dirich- let Allocation(LDA), Non-Negative Matrix Factorization(NMF), Top2Vec, and BERTopic. Topic modelling is an essential method for gaining insights from massive amounts of textual data. Top2Vec and BERTopic are recent approaches that use unsupervised neural networks to develop distributed representations of texts and words, whereas NMF and LDA are traditional techniques frequently utilised for topic modelling. This document gives a timeline of important advances in topic modelling, including the development of NMF and LDA, as well as many refinements and additions to LDA. According to the study’s findings, BERTopic surpasses the other approaches, particularly in recognising overlapping and fine- grained subjects. This work emphasises the significance of text processing quality, the variety of subjects in the text, and the right selection of topic modelling methods in efficiently breaking down topics.

Biomedical Text Classification Using Augmented Word Representation Based on Distributional and Relational Contexts.

Due to the increasing use of information technologies by biomedical experts, researchers, public health agencies, and healthcare professionals, a large number of scientific literatures, clinical notes, and other structured and unstructured text resources are rapidly increasing and being stored in various data sources like PubMed. These massive text resources can be leveraged to extract valuable knowledge and insights using machine learning techniques. Recent advancement in neural network-based classification models has gained popularity which takes numeric vectors (aka word representation) of training data as the input to train classification models. Better the input vectors, more accurate would be the classification. Word representations are learned as the distribution of words in an embedding space, wherein each word has its vector and the semantically similar words based on the contexts appear nearby each other. However, such distributional word representations are incapable of encapsulating relational semantics between distant words. In the biomedical domain, relation mining is a well-studied problem which aims to extract relational words, which associates distant entities generally representing the subject and object of a sentence. Our goal is to capture the relational semantics information between distant words from a large corpus to learn enhanced word representation and employ the learned word representation for various natural language processing tasks such as text classification. In this article, we have proposed an application of biomedical relation triplets to learn word representation through incorporating relational semantic information within the distributional representation of words. In other words, the proposed approach aims to capture both distributional and relational contexts of the words to learn their numeric vectors from text corpus. We have also proposed an application of the learned word representations for text classification. The proposed approach is evaluated over multiple benchmark datasets, and the efficacy of the learned word representations is tested in terms of word similarity and concept categorization tasks. Our proposed approach provides better performance in comparison to the state-of-the-art GloVe model. Furthermore, we have applied the learned word representations to classify biomedical texts using four neural network-based classification models, and the classification accuracy further confirms the effectiveness of the learned word representations by our proposed approach.

Refining Word Embeddings with Sentiment Information for Sentiment Analysis

Natural Language Processing problems generally require the use of pre-trained distributed word representations to be solved with deep learning models. However, distributed representations usually rely on contextual information which prevents them from learning all the important word characteristics. The task of sentiment analysis suffers from such a problem because sentiment information is ignored during the process of learning word embeddings. The performance of sentiment analysis can be affected since two words with similar vectors may have opposite sentiment orientations. The present paper introduces a novel model called Continuous Sentiment Contextualized Vectors (CSCV) to address this problem. The proposed model can learn word sentiment embedding using its surrounding context words. It uses Continuous Bag-of-Words (CBOW) model to deal with the context and sentiment lexicons to identify sentiment. Existing pre-trained vectors are combined then with the obtained sentiment vectors using Principal component analysis (PCA) to enhance their quality. The experiments show that: (1) CSCV vectors can be used to enhance any pre-trained word vectors; (2) The result vectors strongly alleviate the problem of similar words with opposite polarities; (3) The performance of sentiment classification is improved by applying this approach.

LARQS: An Analogical Reasoning Evaluation Dataset for Legal Word Embedding

Applying natural language processing-related algorithms is currently a popular project in legal applications, for instance, document classification of legal documents, contract review and machine translation. Using the above machine learning algorithms, all need to encode the words in the document in the form of vectors. The word embedding model is a modern distributed word representation approach and the most common unsupervised word encoding method. It facilitates subjecting other algorithms and subsequently performing the downstream tasks of natural language processing vis-à-vis. The most common and practical approach of accuracy evaluation with the word embedding model uses a benchmark set with linguistic rules or the relationship between words to perform analogy reasoning via algebraic calculation. This paper proposes establishing a 1,256 Legal Analogical Reasoning Questions Set (LARQS) from the 2,388 Chinese Codex corpus using five kinds of legal relations, which are then used to evaluate the accuracy of the Chinese word embedding model. Moreover, we discovered that legal relations might be ubiquitous in the word embedding model.

Refining electronic medical records representation in manifold subspace

BackgroundElectronic medical records (EMR) contain detailed information about patient health. Developing an effective representation model is of great significance for the downstream applications of EMR. However, processing data directly is difficult because EMR data has such characteristics as incompleteness, unstructure and redundancy. Therefore, preprocess of the original data is the key step of EMR data mining. The classic distributed word representations ignore the geometric feature of the word vectors for the representation of EMR data, which often underestimate the similarities between similar words and overestimate the similarities between distant words. This results in word similarity obtained from embedding models being inconsistent with human judgment and much valuable medical information being lost.ResultsIn this study, we propose a biomedical word embedding framework based on manifold subspace. Our proposed model first obtains the word vector representations of the EMR data, and then re-embeds the word vector in the manifold subspace. We develop an efficient optimization algorithm with neighborhood preserving embedding based on manifold optimization. To verify the algorithm presented in this study, we perform experiments on intrinsic evaluation and external classification tasks, and the experimental results demonstrate its advantages over other baseline methods.ConclusionsManifold learning subspace embedding can enhance the representation of distributed word representations in electronic medical record texts. Reduce the difficulty for researchers to process unstructured electronic medical record text data, which has certain biomedical research value.

Extracting value from Brazilian Court decisions

We propose a methodology to extract value from Brazilian Court decisions to support judges and lawyers in their decision-making. We instantiate our methodology in one information system we have developed. Such system (i) extracts plaintiff’s legal claims and each specific provision on legal opinions enacted by lower and Appellate Courts, and (ii) connects each legal claim with the corresponding judicial provision. The information system presents the results through visualizations. Information Extraction for legal texts has been previously approached in the literature for different languages, using different methods. Our proposal is different from previous work, since our corpora comprise Brazilian lower and Appellate Court decisions, in which we look for a set of plaintiff’s legal claims and judicial provisions commonly judged by the Court. We use the following methods to tackle the information extraction tasks: Bidirectional Long Short-Term Memory network; Conditional Random Fields; and a combination of Bidirectional Long Short-Term Memory network and Conditional Random Fields. In addition to the well-known distributed representation of words in word embeddings, we use character-level representation of words in character embeddings. We have built three corpora – Kauane Insurance Report, Kauane Insurance Lower, and Kauane Insurance Upper – to train and evaluate the system, using public data from the State Court of Rio de Janeiro. Our methods achieved good quality for Kauane Insurance Lower and Kauane Insurance Upper, and promising results for Kauane Insurance Report.

A Scheme of Pairwise Feature Combinations to Improve Sentiment Classification Using Book Review Dataset

Sentiment Analysis is a Natural Language Processing (NLP) domain related to the identification or extraction of user sentiments or opinions from written language. Although the approaches to achieve the goals may vary, Machine Learning (ML) methods are gradually becoming the preferred method because of their ability to automatically draw useful insight from data regardless of their complexity. However, an important prerequisite for most ML algorithms to learn from text data is to encode them into numerical vectors. Popular approaches to this include word level representation methods TF-IDF, distributed word representations (word2vec) and distributed document representations (doc2vec). Each of these methods has demonstrated remarkable success in representing the encoded text, however we found that no method has been set to be excellence in all tasks. Motivated by this challenge, an improved scheme of pairwise fusion are proposed for sentiment classification of book reviews. In the experimental findings, Artificial Neural Networks (ANN) and Logistic Regression (LR) classifiers showed that the proposed scheme improved the performance compared to the single method vectorization method. We see that TF-IDF-word2vec performed best among other methods with a mean accuracy of 91.0% (ANN) and 92.5% (LR); showed an improvement of 0.7% and 0.2% respectively over TF-IDF which is the best single vector method. Thus, the proposed method can used as a compact alternative to the popular bag-of-n-gram models as it captures contextual information of encoded document with a less sparse data.

A Classification Framework of Identifying Major Documents with Search Engine Suggests and Unsupervised Subtopic Clustering

This paper addresses the problem of automatic recognition of out-of-topic documents from a small set of similar documents that are expected to be on some common topic. The objective is to remove documents of noise from a set. A topic model based classification framework is proposed for the task of discovering out-of-topic documents. This paper introduces a new concept of annotated {\it search engine suggests}, where this paper takes whichever search queries were used to search for a page as representations of content in that page. This paper adopted word embedding to create distributed representation of words and documents, and perform similarity comparison on search engine suggests. It is shown that search engine suggests can be highly accurate semantic representations of textual content and demonstrate that our document analysis algorithm using such representation for relevance measure gives satisfactory performance in terms of in-topic content filtering compared to the baseline technique of topic probability ranking.

Learning word representation by jointly using neighbor and syntactic contexts

Interpretability is a significant aspect of the distributed word representation learning model. Although the most advanced pretrained models have achieved the best results till date, the interpretability of a pretrained model is difficult to explain clearly. For this reason, based on the interpretability of distributed word embeddings, this paper presents a method of learning word representation using joint context. At present, the existing distributed word representation models for learning word representations usually focus on either neighbor or syntactic context. We argue that it is necessary to simultaneously model both contexts. In particular, the point mutual information obtained by combining the two types of contexts can efficiently express the correlation between the words. We propose two alternative distribution models for learning word representations by employing the neighbor and syntactic contexts via a simple and effective joint learning framework. Furthermore, the proposed models are trained on a public corpus, and the learned representations are evaluated in word analogy, word similarity, and sentence classification tasks. The experimental results demonstrate the potential of the proposed method.

Effective Training Data Extraction Method to Improve Influenza Outbreak Prediction from Online News Articles: Deep Learning Model Study.

BackgroundEach year, influenza affects 3 to 5 million people and causes 290,000 to 650,000 fatalities worldwide. To reduce the fatalities caused by influenza, several countries have established influenza surveillance systems to collect early warning data. However, proper and timely warnings are hindered by a 1- to 2-week delay between the actual disease outbreaks and the publication of surveillance data. To address the issue, novel methods for influenza surveillance and prediction using real-time internet data (such as search queries, microblogging, and news) have been proposed. Some of the currently popular approaches extract online data and use machine learning to predict influenza occurrences in a classification mode. However, many of these methods extract training data subjectively, and it is difficult to capture the latent characteristics of the data correctly. There is a critical need to devise new approaches that focus on extracting training data by reflecting the latent characteristics of the data.ObjectiveIn this paper, we propose an effective method to extract training data in a manner that reflects the hidden features and improves the performance by filtering and selecting only the keywords related to influenza before the prediction.MethodsAlthough word embedding provides a distributed representation of words by encoding the hidden relationships between various tokens, we enhanced the word embeddings by selecting keywords related to the influenza outbreak and sorting the extracted keywords using the Pearson correlation coefficient in order to solely keep the tokens with high correlation with the actual influenza outbreak. The keyword extraction process was followed by a predictive model based on long short-term memory that predicts the influenza outbreak. To assess the performance of the proposed predictive model, we used and compared a variety of word embedding techniques.ResultsWord embedding without our proposed sorting process showed 0.8705 prediction accuracy when 50.2 keywords were selected on average. Conversely, word embedding using our proposed sorting process showed 0.8868 prediction accuracy and an improvement in prediction accuracy of 12.6%, although smaller amounts of training data were selected, with only 20.6 keywords on average.ConclusionsThe sorting stage empowers the embedding process, which improves the feature extraction process because it acts as a knowledge base for the prediction component. The model outperformed other current approaches that use flat extraction before prediction.

A Classification Framework of Identifying Major Documents With Search Engine Suggestions and Unsupervised Subtopic Clustering

This paper addresses the problem of automatic recognition of out-of-topic documents from a small set of similar documents that are expected to be on some common topic. The objective is to remove documents of noise from a set. A topic model based classification framework is proposed for the task of discovering out-of-topic documents. This paper introduces a new concept of annotated {\it search engine suggests}, where this paper takes whichever search queries were used to search for a page as representations of content in that page. This paper adopted word embedding to create distributed representation of words and documents, and perform similarity comparison on search engine suggests. It is shown that search engine suggests can be highly accurate semantic representations of textual content and demonstrate that our document analysis algorithm using such representation for relevance measure gives satisfactory performance in terms of in-topic content filtering compared to the baseline technique of topic probability ranking.

Pre-training Context and Time Aware Location Embeddings from Spatial-Temporal Trajectories for User Next Location Prediction

Pre-training location embeddings from spatial-temporal trajectories is a fundamental procedure and very beneficial for user next location prediction. In the real world, a location usually has variable functionalities under different contextual environments. If the exact functions of a location in the trajectory can be reflected in its embedding, the accuracy of user next location prediction should be improved. Yet, existing location embeddings pre-trained on trajectories are mostly based on distributed word representations, which mix a location's various functionalities into one latent representation vector. To address this problem, we propose a Context and Time aware Location Embedding (CTLE) model, which calculates a location's representation vector with consideration of its specific contextual neighbors in trajectories. In this way, the multi-functional properties of locations can be properly tackled. Furthermore, in order to incorporate temporal information in trajectories into location embeddings, we propose a subtle temporal encoding module and a novel pre-training objective, which further improve the quality of location embeddings. We evaluate our proposed model on two real-world mobile user trajectory datasets. The experimental results demonstrate that, compared with the existing embedding methods, our CTLE model can pre-train higher quality location embeddings and significantly improve the performance of downstream user location prediction models.

Probing Lexical Ambiguity: Word Vectors Encode Number and Relatedness of Senses.

Lexical ambiguity-the phenomenon of a single word having multiple, distinguishable senses-is pervasive in language. Both the degree of ambiguity of a word (roughly, its number of senses) and the relatedness of those senses have been found to have widespread effects on language acquisition and processing. Recently, distributional approaches to semantics, in which a word's meaning is determined by its contexts, have led to successful research quantifying the degree of ambiguity, but these measures have not distinguished between the ambiguity of words with multiple related senses versus multiple unrelated meanings. In this work, we present the first assessment of whether distributional meaning representations can capture the ambiguity structure of a word, including both the number and relatedness of senses. On a very large sample of English words, we find that some, but not all, distributional semantic representations that we test exhibit detectable differences between sets of monosemes (unambiguous words; N=964), polysemes (with multiple related senses; N=4,096), and homonyms (with multiple unrelated senses; N=355). Our findings begin to answer open questions from earlier work regarding whether distributional semantic representations of words, which successfully capture various semantic relationships, also reflect fine-grained aspects of meaning structure that influence human behavior. Our findings emphasize the importance of measuring whether proposed lexical representations capture such distinctions: In addition to standard benchmarks that test the similarity structure of distributional semantic models, we need to also consider whether they have cognitively plausible ambiguity structure.

Efficient estimation of Hindi WSD with distributed word representation in vector space

Word Sense Disambiguation (WSD) is significant for improving the accuracy of the interpretation of a Natural language text. Various supervised learning-based models and knowledge-based models have been developed in the literature for WSD of the language text. However, these models do not provide good results for low-resource languages, due to the lack of labelled and tagged data. Therefore, in this work, we have examined different word embedding techniques for word sense disambiguation of the Hindi language texts. Several studies in the literature show that these embeddings have been utilized for different foreign languages in the field of word sense disambiguation. However, to the best of our knowledge, no such work exists for the Hindi language. Therefore, in this paper, we utilize various existing word embeddings for WSD of Hindi text. Moreover, we have created Hindi word embeddings on articles taken from Wikipedia and test the quality of the created word embeddings using Pearson correlation. In this direction, we perform different experiments and observe that Word2Vec model gives best performance among all the considered embeddings on the used Hindi dataset. In our method, the proposed model directly takes input that is trained with word embedding methods and helps to develop a sense inventory using clustering that has been employed for performing disambiguation. Experimental observations indicate that the performance of the proposed approach is moderate and competent in terms of accuracy. The paper, thus, presents how WSD can leverage these representations to encode rich semantic information.

Improving biomedical word representation with locally linear embedding

Distributed word representation, usually obtained through calculation from large corpora, has been widely used in biomedical text because of its effectiveness in representing word semantic information. High-quality and meaningful biomedical words enable doctors to obtain the gist of information and knowledge in a short time to make clinical decisions quickly. Currently, the distributed word representation ignores the influence of the word embedding geometric structure obtained through calculation on the word semantic information and cannot accurately represent the word information, thus affecting the representation effect of biomedical text. To solve the above problems, we propose a biomedical word embedding framework based on manifold learning. Our work provides new perspectives for representing biomedical word embedding, which is the key concept in biomedical natural language processing tasks. First, the distributed word representation model is used to obtain the pretrained word embedding, and then the manifold learning is used to re-embed the pretrained word embedding. To verify the validity of the proposed framework in the biomedical domain, we evaluate the algorithm by using biomedical texts. Experimental results show that the proposed method can effectively improve the results of electronic health record classification and semantic similarity.

Word Embeddings for Constructive Comments Classification

Word embeddings are so significant today that it is common to see their application in multiple natural language tasks. Indeed, word embeddings as the first layer of a deep learning model are widely adopted and they can also be found in multiple natural language tasks such as classification of texts and named entity recognition. The focus of this paper is the identification of constructive online comments through the use of dense vector semantics such as word embeddings. We specifically explore two approaches: learning distributed word representations and the use of a pre-trained word embedding model. We evaluate these word embedding methods on a recently created constructive comments corpus comprised of 12,000 annotated news comments, intended to improve the quality of online discussions. The obtained results show how the performance of complicated architectures like recurrent and convolutional neural networks can be matched by a language model based on learning embeddings.

Distributional Word Representations for Code-mixed Text in Moroccan Darija

In this paper, we build different word embedding representations for code-mixed Moroccan Darija-French-English text. We begin by collecting a large corpus based on social media posts and song lyrics. We then experiment with four algorithms: namely, Word2Vec using both Continuous Bag of Words (CBOW) and SkipGram flavors, FastText-CBOW and FastText-SkipGram. Finally, we evaluate these models on two tasks: Word Analogies and Word-Level Language Detection. The use of character n-grams in the FastText models allowed them to perform much better than their word2vec counterparts on the word analogy task, while Word2Vec-SkipGram performed best on the language identification. To the knowledge of the authors, our embeddings are the first set of distributed representations for Arabizi Moroccan Darija text. In the context of this work, we also managed to build the first ever code-switched English – French - Moroccan Darija word analogy dataset and the largest Arabizi code-switched English – French - Moroccan Darija.

Example Phrase Adaptation Method for Customized, Example-Based Dialog System Using User Data and Distributed Word Representations

Example Phrase Adaptation Method for Customized, Example-Based Dialog System Using User Data and Distributed Word Representations

Unsupervised Predominant Sense Detection and Its Application to Text Classification

This paper focuses on the domain-specific senses of words and proposes a method for detecting predominant sense depending on each domain. Our Domain-Specific Senses (DSS) model is an unsupervised manner and detects predominant senses in each domain. We apply a simple Markov Random Walk (MRW) model to ranking senses for each domain. It decides the importance of a sense within a graph by using the similarity of senses. The similarity of senses is obtained by using distributional representations of words from gloss texts in the thesaurus. It can capture large semantic context and thus does not require manual annotation of sense-tagged data. We used the Reuters corpus and the WordNet in the experiments. We applied the results of domain-specific senses to text classification and examined how DSS affects the overall performance of the text classification task. We compared our DSS model with one of the word sense disambiguation techniques (WSD), Context2vec, and the results demonstrate our domain-specific sense approach gains 0.053 F1 improvement on average over the WSD approach.