Distributional Word Embeddings Research Articles

Improving Domain-Generalized Few-Shot Text Classification with Multi-Level Distributional Signatures

Domain-generalized few-shot text classification (DG-FSTC) is a new setting for few-shot text classification (FSTC). In DG-FSTC, the model is meta-trained on a multi-domain dataset, and meta-tested on unseen datasets with different domains. However, previous methods mostly construct semantic representations by learning from words directly, which is limited in domain adaptability. In this study, we enhance the domain adaptability of the model by utilizing the distributional signatures of texts that indicate domain-related features in specific domains. We propose a Multi-level Distributional Signatures based model, namely MultiDS. Firstly, inspired by pretrained language models, we compute distributional signatures from an extra large news corpus, and we denote these as domain-agnostic features. Then we calculate the distributional signatures from texts in the same domain and texts from the same class, respectively. These two kinds of information are regarded as domain-specific and class-specific features, respectively. After that, we fuse and translate these three distributional signatures into word-level attention values, which enables the model to capture informative features as domain changes. In addition, we utilize domain-specific distributional signatures for the calibration of feature representations in specific domains. The calibration vectors produced by the domain-specific distributional signatures and word embeddings help the model adapt to various domains. Extensive experiments are performed on four benchmarks. The results demonstrate that our proposed method beats the state-of-the-art method with an average improvement of 1.41% on four datasets. Compared with five competitive baselines, our method achieves the best average performance. The ablation studies prove the effectiveness of each proposed module.

Early prediction of student knowledge in game‐based learning with distributed representations of assessment questions

AbstractGame‐based learning environments hold significant promise for facilitating learning experiences that are both effective and engaging. To support individualised learning and support proactive scaffolding when students are struggling, game‐based learning environments should be able to accurately predict student knowledge at early points in students' gameplay. Student knowledge is traditionally assessed prior to and after each student interacts with the learning environment with conventional methods, such as multiple choice content knowledge assessments. While previous student modelling approaches have leveraged machine learning to automatically infer students' knowledge, there is limited work that incorporates the fine‐grained content from each question in these types of tests into student models that predict student performance at early junctures in gameplay episodes. This work investigates a predictive student modelling approach that leverages the natural language text of the post‐gameplay content knowledge questions and the text of the possible answer choices for early prediction of fine‐grained individual student performance in game‐based learning environments. With data from a study involving 66 undergraduate students from a large public university interacting with a game‐based learning environment for microbiology, Crystal Island, we investigate the accuracy and early prediction capacity of student models that use a combination of gameplay features extracted from student log files as well as distributed representations of post‐test content assessment questions. The results demonstrate that by incorporating knowledge about assessment questions, early prediction models are able to outperform competing baselines that only use student game trace data with no question‐related information. Furthermore, this approach achieves high generalisation, including predicting the performance of students on unseen questions. Practitioner notesWhat is already known about this topic A distinctive characteristic of game‐based learning environments is their capacity to enable fine‐grained student assessment. Adaptive game‐based learning environments offer individualisation based on specific student needs and should be able to assess student competencies using early prediction models of those competencies. Word embedding approaches from the field of natural language processing show great promise in the ability to encode semantic information that can be leveraged by predictive student models. What this paper adds Investigates word embeddings of assessment question content for reliable early prediction of student performance. Demonstrates the efficacy of distributed word embeddings of assessment questions when used by early prediction models compared to models that use either no assessment information or discrete representations of the questions. Demonstrates the efficacy and generalisability of word embeddings of assessment questions for predicting the performance of both new students on existing questions and existing students on new questions. Implications for practice and/or policy Word embeddings of assessment questions can enhance early prediction models of student knowledge, which can drive adaptive feedback to students who interact with game‐based learning environments. Practitioners should determine if new assessment questions will be developed for their game‐based learning environment, and if so, consider using our student modelling framework that incorporates early prediction models pretrained with existing student responses to previous assessment questions and is generalisable to the new assessment questions by leveraging distributed word embedding techniques. Researchers should consider the most appropriate way to encode the assessment questions in ways that early prediction models are able to infer relationships between the questions and gameplay behaviour to make accurate predictions of student competencies.

Improving Zero-Shot Learning Baselines with Commonsense Knowledge

Zero-shot learning — the problem of training and testing on a completely disjoint set of classes — relies greatly on its ability to transfer knowledge from train classes to test classes. Traditionally semantic embeddings consisting of human-defined attributes or distributed word embeddings are used to facilitate this transfer by improving the association between visual and semantic embeddings. In this paper, we take advantage of explicit relations between nodes defined in ConceptNet, a commonsense knowledge graph, to generate commonsense embeddings of the class labels by using a graph convolution network-based autoencoder. Our experiments performed on three standard benchmark datasets surpass the strong baselines when we fuse our commonsense embeddings with existing semantic embeddings, i.e., human-defined attributes and distributed word embeddings. This work paves the path to more brain-inspired approaches to zero-short learning.

Evaluating Biomedical Word Embeddings for Vocabulary Alignment at Scale in the UMLS Metathesaurus Using Siamese Networks.

Recent work uses a Siamese Network, initialized with BioWordVec embeddings (distributed word embeddings), for predicting synonymy among biomedical terms to automate a part of the UMLS (Unified Medical Language System) Metathesaurus construction process. We evaluate the use of contextualized word embeddings extracted from nine different biomedical BERT-based models for synonymy prediction in the UMLS by replacing BioWordVec embeddings with embeddings extracted from each biomedical BERT model using different feature extraction methods. Surprisingly, we find that Siamese Networks initialized with BioWordVec embeddings still outperform the Siamese Networks initialized with embedding extracted from biomedical BERT model.

Categorizing Sexism and Misogyny through Neural Approaches

Sexism, an injustice that subjects women and girls to enormous suffering, manifests in blatant as well as subtle ways. In the wake of growing documentation of experiences of sexism on the web, the automatic categorization of accounts of sexism has the potential to assist social scientists and policymakers in studying and thereby countering sexism. The existing work on sexism classification has certain limitations in terms of the categories of sexism used and/or whether they can co-occur. To the best of our knowledge, this is the first work on the multi-label classification of sexism of any kind(s). 1 We also consider the related task of misogyny classification. While sexism classification is performed on textual accounts describing sexism suffered or observed, misogyny classification is carried out on tweets perpetrating misogyny. We devise a novel neural framework for classifying sexism and misogyny that can combine text representations obtained using models such as Bidirectional Encoder Representations from Transformers with distributional and linguistic word embeddings using a flexible architecture involving recurrent components and optional convolutional ones. Further, we leverage unlabeled accounts of sexism to infuse domain-specific elements into our framework. To evaluate the versatility of our neural approach for tasks pertaining to sexism and misogyny, we experiment with adapting it for misogyny identification. For categorizing sexism, we investigate multiple loss functions and problem transformation techniques to address the multi-label problem formulation. We develop an ensemble approach using a proposed multi-label classification model with potentially overlapping subsets of the category set. Proposed methods outperform several deep-learning as well as traditional machine learning baselines for all three tasks.

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.

BnVec: Towards the Development of Word Embedding for Bangla Language Processing

Progression in machine learning and statistical inference are facilitating the advancement of domains like computer vision, natural language processing (NLP), automation & robotics, and so on. Among the different persuasive improvements in NLP, word embedding is one of the most used and revolutionary techniques. In this paper, we manifest an open-source library for Bangla word extraction systems named BnVec which expects to furnish the Bangla NLP research community by the utilization of some incredible word embedding techniques. The BnVec is splitted up into two parts, the first one is the Bangla suitable defined class to embed words with access to the six most popular word embedding schemes (CountVectorizer, TF-IDF, Hash Vectorizer, Word2vec, fastText, and Glove). The other one is based on the pre-trained distributed word embedding system of Word2vec, fastText, and GloVe. The pre-trained models have been built by collecting content from the newspaper, social media, and Bangla wiki articles. The total number of tokens used to build the models exceeds 395,289,960. The paper additionally depicts the performance of these models by various hyper-parameter tuning and then analyzes the results.

Learning Lexical Subspaces in a Distributional Vector Space

In this paper, we propose LexSub, a novel approach towards unifying lexical and distributional semantics. We inject knowledge about lexical-semantic relations into distributional word embeddings by defining subspaces of the distributional vector space in which a lexical relation should hold. Our framework can handle symmetric attract and repel relations (e.g., synonymy and antonymy, respectively), as well as asymmetric relations (e.g., hypernymy and meronomy). In a suite of intrinsic benchmarks, we show that our model outperforms previous approaches on relatedness tasks and on hypernymy classification and detection, while being competitive on word similarity tasks. It also outperforms previous systems on extrinsic classification tasks that benefit from exploiting lexical relational cues. We perform a series of analyses to understand the behaviors of our model. 1 Code available at https://github.com/aishikchakraborty/LexSub .

Better synonyms for enriching biomedical search.

In a biomedical literature search, the link between a query and a document is often not established, because they use different terms to refer to the same concept. Distributional word embeddings are frequently used for detecting related words by computing the cosine similarity between them. However, previous research has not established either the best embedding methods for detecting synonyms among related word pairs or how effective such methods may be. In this study, we first create the BioSearchSyn set, a manually annotated set of synonyms, to assess and compare 3 widely used word-embedding methods (word2vec, fastText, and GloVe) in their ability to detect synonyms among related pairs of words. We demonstrate the shortcomings of the cosine similarity score between word embeddings for this task: the same scores have very different meanings for the different methods. To address the problem, we propose utilizing pool adjacent violators (PAV), an isotonic regression algorithm, to transform a cosine similarity into a probability of 2 words being synonyms. Experimental results using the BioSearchSyn set as a gold standard reveal which embedding methods have the best performance in identifying synonym pairs. The BioSearchSyn set also allows converting cosine similarity scores into probabilities, which provides a uniform interpretation of the synonymy score over different methods. We introduced the BioSearchSyn corpus of 1000 term pairs, which allowed us to identify the best embedding method for detecting synonymy for biomedical search. Using the proposed method, we created PubTermVariants2.0: a large, automatically extracted set of synonym pairs that have augmented PubMed searches since the spring of 2019.

Deep learning based emotion analysis of microblog texts

Traditional text emotion analysis methods are primarily devoted to studying extended texts, such as news reports and full-length documents. Microblogs are considered short texts that are often characterized by large noises, new words, and abbreviations. Previous emotion classification methods usually fail to extract significant features and achieve poor classification effect when applied to processing of short texts or micro-texts. This study proposes a microblog emotion classification model, namely, CNN_Text_Word2vec, on the basis of convolutional neural network (CNN) to solve the above-mentioned problems. CNN_Text_Word2vec introduces a word2vec neural network model to train distributed word embeddings on every single word. The trained word vectors are used as input features for the model to learn microblog text features through parallel convolution layers with multiple convolution kernels of different sizes. Experiment results show that the overall accuracy rate of CNN_Text_Word2vec is 7.0% higher than that achieved by current mainstream methods, such as SVM, LSTM and RNN. Moreover, this study explores the impact of different semantic units on the accuracy of CNN_Text_Word2vec, specifically in processing of Chinese texts. The experimental results show that comparing to using feature vectors obtained from training words, feature vector obtained from training Chinese characters yields a better performance.

Exploring What Is Encoded in Distributional Word Vectors: A Neurobiologically Motivated Analysis

The pervasive use of distributional semantic models or word embeddings for both cognitive modeling and practical application is because of their remarkable ability to represent the meanings of words. However, relatively little effort has been made to explore what types of information are encoded in distributional word vectors. Knowing the internal knowledge embedded in word vectors is important for cognitive modeling using distributional semantic models. Therefore, in this paper, we attempt to identify the knowledge encoded in word vectors by conducting a computational experiment using Binder et al.'s (2016) featural conceptual representations based on neurobiologically motivated attributes. In an experiment, these conceptual vectors are predicted from text-based word vectors using a neural network and linear transformation, and prediction performance is compared among various types of information. The analysis demonstrates that abstract information is generally predicted more accurately by word vectors than perceptual and spatiotemporal information, and specifically, the prediction accuracy of cognitive and social information is higher. Emotional information is also found to be successfully predicted for abstract words. These results indicate that language can be a major source of knowledge about abstract attributes, and theysupport the recent view that emphasizes the importance of language for abstract concepts. Furthermore, we show that word vectors can capture some types of perceptual and spatiotemporal information about concrete concepts and some relevant word categories. This suggests that language statistics can encode more perceptual knowledge than often expected.

Improving Chinese Word Representation with Conceptual Semantics

The meaning of a word includes a conceptual meaning and a distributive meaning. Word embedding based on distribution suffers from insufficient conceptual semantic representation caused by data sparsity, especially for low-frequency words. In knowledge bases, manually annotated semantic knowledge is stable and the essential attributes of words are accurately denoted. In this paper, we propose a Conceptual Semantics Enhanced Word Representation (CEWR) model, computing the synset embedding and hypernym embedding of Chinese words based on the Tongyici Cilin thesaurus, and aggregating it with distributed word representation to have both distributed information and the conceptual meaning encoded in the representation of words. We evaluate the CEWR model on two tasks: word similarity computation and short text classification. The Spearman correlation between model results and human judgement are improved to 64.71%, 81.84%, and 85.16% on Wordsim297, MC30, and RG65, respectively. Moreover, CEWR improves the F1 score by 3% in the short text classification task. The experimental results show that CEWR can represent words in a more informative approach than distributed word embedding. This proves that conceptual semantics, especially hypernymous information, is a good complement to distributed word representation.

Extracting Word Synonyms from Text using Neural Approaches

Extracting synonyms from textual corpora using computational techniques is an interesting research problem in the Natural Language Processing (NLP) domain. Neural techniques (such as Word2Vec) have been recently utilized to produce distributional word representations (also known as word embeddings) that capture semantic similarity/relatedness between words based on linear context. Nevertheless, using these techniques for synonyms extraction poses many challenges due to the fact that similarity between vector word representations does not indicate only synonymy between words, but also other sense relations as well as word association or relatedness. In this paper, we tackle this problem using a novel 2-step approach. We first build distributional word embeddings using Word2Vec then use the induced word embeddings as an input to train a feed-forward neutral network using annotated dataset to distinguish between synonyms and other semantically related words

More than Bags of Words: Sentiment Analysis with Word Embeddings

ABSTRACTMoving beyond the dominant bag-of-words approach to sentiment analysis we introduce an alternative procedure based on distributed word embeddings. The strength of word embeddings is the ability to capture similarities in word meaning. We use word embeddings as part of a supervised machine learning procedure which estimates levels of negativity in parliamentary speeches. The procedure’s accuracy is evaluated with crowdcoded training sentences; its external validity through a study of patterns of negativity in Austrian parliamentary speeches. The results show the potential of the word embeddings approach for sentiment analysis in the social sciences.

Two layers LSTM with attention for multi-choice question answering in exams

Question Answering in Exams is typical question answering task that aims to test how accurately the model could answer the questions in exams. In this paper, we use general deep learning model to solve the multi-choice question answering task. Our approach is to build distributed word embedding of question and answers instead of manually extracting features or linguistic tools, meanwhile, for improving the accuracy, the external corpus is introduced. The framework uses a two layers LSTM with attention which get a significant result. By contrast, we introduce the simple long short-term memory (QA-LSTM) model and QA-LSTM-CNN model and QA-LSTM with attention model as the reference. Experiment demonstrate superior performance of two layers LSTM with attention compared to other models in question answering task.

Extended Distributed Prototypical For Biomedical Named Entity Recognition

Biomedical Named Entity Recognition (Bio-NER) is an essential step of biomedical information extraction and biomedical text mining. Although, a lot of researches have been made in the design of rule-based and supervised tools for general NER, Bio-NER still remains a challenge and an area of active research, as still there is huge difference in F-score of 10 points between general newswire NER and Bio-NER. The complex structures of the biomedical entities pose a huge challenge for their recognition. To handle this, this paper explores different effective word representations with Support Vector Machine (SVM) to deal with the complex structures of biomedical named entities. First, this paper identifies and evaluates a set of morphological and contextual features with SVM learning method for Bio-NER. This paper also presents an extended distributed representation word embedding technique (EDRWE) for Bio-NER. These models are evaluated on widely used standard Bio-NER dataset namely GENIA corpus. Experimental results show that EDRWE technique improves the overall performance of the Bio-NER and outperforms all other representation methods. Results analysis shows that the new EDRWE is satisfactory and effective for Bio-NER especially when only a small-sized data set is available.

Survey on distributed word embeddings based on neural network language models

Survey on distributed word embeddings based on neural network language models

Learning linear transformations between counting-based and prediction-based word embeddings.

Despite the growing interest in prediction-based word embedding learning methods, it remains unclear as to how the vector spaces learnt by the prediction-based methods differ from that of the counting-based methods, or whether one can be transformed into the other. To study the relationship between counting-based and prediction-based embeddings, we propose a method for learning a linear transformation between two given sets of word embeddings. Our proposal contributes to the word embedding learning research in three ways: (a) we propose an efficient method to learn a linear transformation between two sets of word embeddings, (b) using the transformation learnt in (a), we empirically show that it is possible to predict distributed word embeddings for novel unseen words, and (c) empirically it is possible to linearly transform counting-based embeddings to prediction-based embeddings, for frequent words, different POS categories, and varying degrees of ambiguities.

Decoding naturalistic experiences from human brain activity via distributed representations of words

Natural visual scenes induce rich perceptual experiences that are highly diverse from scene to scene and from person to person. Here, we propose a new framework for decoding such experiences using a distributed representation of words. We used functional magnetic resonance imaging (fMRI) to measure brain activity evoked by natural movie scenes. Then, we constructed a high-dimensional feature space of perceptual experiences using skip-gram, a state-of-the-art distributed word embedding model. We built a decoder that associates brain activity with perceptual experiences via the distributed word representation. The decoder successfully estimated perceptual contents consistent with the scene descriptions by multiple annotators. Our results illustrate three advantages of our decoding framework: (1) three types of perceptual contents could be decoded in the form of nouns (objects), verbs (actions), and adjectives (impressions) contained in 10,000 vocabulary words; (2) despite using such a large vocabulary, we could decode novel words that were absent in the datasets to train the decoder; and (3) the inter-individual variability of the decoded contents co-varied with that of the contents of scene descriptions. These findings suggest that our decoding framework can recover diverse aspects of perceptual experiences in naturalistic situations and could be useful in various scientific and practical applications.

Leveraging semantic resources in diversified query expansion

A search query, being a very concise grounding of user intent, could potentially have many possible interpretations. Search engines hedge their bets by diversifying top results to cover multiple such possibilities so that the user is likely to be satisfied, whatever be her intended interpretation. Diversified Query Expansion is the problem of diversifying query expansion suggestions, so that the user can specialize the query to better suit her intent, even before perusing search results. In this paper, we consider the usage of semantic resources and tools to arrive at improved methods for diversified query expansion. In particular, we develop two methods, those that leverage Wikipedia and pre-learnt distributional word embeddings respectively. Both the approaches operate on a common three-phase framework; that of first taking a set of informative terms from the search results of the initial query, then building a graph, following by using a diversity-conscious node ranking to prioritize candidate terms for diversified query expansion. Our methods differ in the second phase, with the first method Select-Link-Rank (SLR) linking terms with Wikipedia entities to accomplish graph construction; on the other hand, our second method, Select-Embed-Rank (SER), constructs the graph using similarities between distributional word embeddings. Through an empirical analysis and user study, we show that SLR ourperforms state-of-the-art diversified query expansion methods, thus establishing that Wikipedia is an effective resource to aid diversified query expansion. Our empirical analysis also illustrates that SER outperforms the baselines convincingly, asserting that it is the best available method for those cases where SLR is not applicable; these include narrow-focus search systems where a relevant knowledge base is unavailable. Our SLR method is also seen to outperform a state-of-the-art method in the task of diversified entity ranking.