Unsupervised Post-Processing of Word Vectors via Conceptor Negation

With growing applications of Machine Learning in daily lives Natural Language Processing (NLP) has emerged as a heavily researched area. Finding its applications in tasks ranging from simple Q/A chatbots to Fully fledged conversational AI, NLP models are vital. Word and Sentence embedding are one of the most common starting points of any NLP task. A word embedding represents a given word in a predefined vector-space while maintaining vector relations with similar or dis-similar entities. As such different pretrained embedding such as Word2Vec, GloVe, fasttext have been developed. These embedding generated on millions of words are however very large in terms of size. Having embedding with floating point precision also makes the downstream evaluation slow. In this paper we present a novel method to convert continuous embedding to its binary representation, thus reducing the overall size of the embedding while keeping the semantic and relational knowledge intact. This will facilitate an option of porting such big embedding onto devices where space is limited. We also present different approaches suitable for different downstream tasks based on the requirement of contextual and semantic information. Experiments have shown comparable result in downstream tasks with 7 to 15 times reduction in file size and about 5 % change in evaluation parameters.

Word embeddings based on a conditional model are commonly used in Natural Language Processing (NLP) tasks to embed the words of a dictionary in a low dimensional linear space. Their computation is based on the maximization of the likelihood of a conditional probability distribution for each word of the dictionary. These distributions form a Riemannian statistical manifold, where word embeddings can be interpreted as vectors in the tangent space of a specific reference measure on the manifold. A novel family of word embeddings, called -embeddings have been recently introduced as deriving from the geometrical deformation of the simplex of probabilities through a parameter , using notions from Information Geometry. After introducing the -embeddings, we show how the deformation of the simplex, controlled by , provides an extra handle to increase the performances of several intrinsic and extrinsic tasks in NLP. We test the -embeddings on different tasks with models of increasing complexity, showing that the advantages associated with the use of -embeddings are present also for models with a large number of parameters. Finally, we show that tuning allows for higher performances compared to the use of larger models in which additionally a transformation of the embeddings is learned during training, as experimentally verified in attention models.

Text classification is an important and classic task of natural language processing. Deep neural networks are becoming more and more popular in text classification due to their expressive power and low requirements for feature engineering. However, the more flexible graph convolutional neural network is rarely used for this task. This paper proposes a text classification model (GENET) based on graph convolution word embedding and attention mechanism. The model can better combine the semantic and lexical information of the text with the discontinuous global word co-occurrence information and long-distance semantic information in the corpus. It breaks the shortcomings of traditional neural networks that have limited structure and can only learn local information. Experimental results show that our proposed text classification method out performs other models on multiple datasets. At the same time, it is proved through experiments that the word global information obtained by GCN is an important supplement to the word embedding representation.

Learning the inherent meaning of a word in Natural Language Processing (NLP) has motivated researchers to represent a word at various levels of abstraction, namely character-level, morpheme-level, and subword-level vector representations. Syllable-Aware Word Embedding (SAWE) can effectively handle agglutinative and fusion-based NLP tasks. However, research attempts on assessing the SAWE on such extrinsic NLP tasks has been scanty, especially for low-resource languages in the context of code-mixing with English. A model to learn SAWE to extract semantics at fine-grained subunits of a word is proposed in this article, and the representative ability of the embeddings is assessed through sentiment analysis of code-mixed Telugu-English review corpora. Multilingual societies and advancements in communication technologies have accounted for the prolific usage of mixed data, which renders the State-of-the-Art (SOTA) sentiment analysis models developed based on monolingual data ineffective. Social media users in the Indian subcontinent exhibit a tendency to mix English and their respective native language (using the phonetic form of English) in expressing their opinions or sentiments. A code-mixing scenario provides flexibility to borrow words from a foreign language, usage of shorthand notations, elongation of vowels, and usage of words without following syntactic/grammatical rules, which renders the sentiment analysis of code-mixed data challenging to perform. Deep neural architectures like Long Short-Term Memory and Gated Recurrent Unit networks have been shown to be effective in solving several NLP tasks, such as sequence labeling, named entity recognition, and machine translation. In this article, a framework to perform sentiment analysis on a code-mixed Telugu-English review corpus is implemented. Both word embedding and SAWE are input to a unified deep neural network that contains a two-level Bidirectional Long Short-Term Memory/Gated Recurrent Unit network with Softmax as the output layer. The proposed model leverages the advantages of both word embedding and SAWE, which enable the proposed model to outperform existing SOTA code-mixed sentiment analysis models on a Telugu-English code-mixed dataset of the International Institute of Information Technology–Hyderabad and a dataset curated by the authors. The improvement realized by the proposed model on these datasets is [3% increase in F1-score and 2% increase in accuracy] and [7% increase in F1-score and 5% in accuracy], respectively, in comparison with the best-performing SOTA model.

Word embeddings, distributed word representations in a reduced linear space, show a lot of promise for accomplishing Natural Language Processing (NLP) tasks in an unsupervised manner. In this study, we investigate if the success of word2vec, a Neural Networks based word embeddings algorithm, can be replicated in an aggluginative language like Turkish. Turkish is more challenging than languages like English for complex NLP tasks because of her rich morphology. We picked ontology enrichment, again a relatively harder NLP task, as our test application. Firstly, we show how ontological relations can be extracted automaticaly from Turkish Wikipedia to construct a gold standard. Then by running experiments we show that the word vector representations produced by word2vec are useful to detect ontological relations encoded in Wikipedia. We propose a simple but yet effective weakly supervised ontology enrichment algorithm where for a given word a few know ontologically related concepts coupled with similarity scores computed via word2vec models can result in discovery of other related concepts. We argue how our algorithm can be improved and augmented to make it a viable component of an ontoloy learning and population framework.

Distributed representations of words have been an indispensable component for natural language processing (NLP) tasks. However, the large memory footprint of word embeddings makes it challenging to deploy NLP models to memory-constrained devices (e.g., self-driving cars, mobile devices). In this paper, we propose a novel method to adaptively compress word embeddings. We fundamentally follow a code-book approach that represents words as discrete codes such as (8, 5, 2, 4). However, unlike prior works that assign the same length of codes to all words, we adaptively assign different lengths of codes to each word by learning downstream tasks. The proposed method works in two steps. First, each word directly learns to select its code length in an end-to-end manner by applying the Gumbel-softmax tricks. After selecting the code length, each word learns discrete codes through a neural network with a binary constraint. To showcase the general applicability of the proposed method, we evaluate the performance on four different downstream tasks. Comprehensive evaluation results clearly show that our method is effective and makes the highly compressed word embeddings without hurting the task accuracy. Moreover, we show that our model assigns word to each code-book by considering the significance of tasks.

Compared with traditional methods, word em-bedding is an efficient language representation that can learn syntax and semantics by using neural networks. As the result, more and more promising experiments in natural language processing (NLP) get the state-of-the-art results by introducing word embedding. In principle, embedding representation learning embeds words to a low-dimensional vector space, there-fore vectors support initialization of NLP tasks such as text classification, sentiment analysis, language understanding, etc. However, polysemy is very common in many languages, which causes word ambiguation, further influences the accuracy of the system. Additionally, language models based on distributed hypotheses mostly focused on word properties rather than morphology were our primary focus. This leads to unreasonable performance in different evaluations. At the same time, word embedding learning and measuring are two vital components of word representation. In this paper, we overviewed many language models including single sense and multiple sense word embedding, and many evaluated approaches including intrinsic and extrinsic evaluation. We found that there are obvious gaps between vectors and manual annotations in word similarity evaluation, and language models that achieved good performance in intrinsic evaluations could not produce similar results in extrinsic evaluations. To the best of our knowledge, there is no universal language model and embedding learning method for most NLP task, and each evaluations also hidden natural defects compared to human knowledge. More evaluated datasets are also investigated such as datasets used in intrinsic and extrinsic evaluations. We believe that an improved evaluation dataset and a more rational evaluation method would benefit from this overview.

SummaryTaxonomic relations play an important role in various Natural Language Processing (NLP) tasks (eg, information extraction, question answering and knowledge inference). Existing approaches on embedding‐based taxonomic relation learning mainly rely on the word embeddings trained using co‐occurrence‐based similarity learning. However, the performance of these approaches is not quite satisfactory due to the lack of sufficient taxonomic semantic knowledge within word embeddings. To solve this problem, we propose an improved embedding‐based approach to learn taxonomic relations via incorporating relation descriptions into word embeddings. First, to capture additional taxonomic semantic knowledge, we train special word embeddings using not only co‐occurrence information of words but also relation descriptions (eg, taxonomic seed relations and their contextual triples). Then, using the trained word embeddings as features, we employ two learning models to identify and predict taxonomic relations, namely, offset‐based classification model and offset‐based similarity model. Experimental results on four real‐world domain datasets demonstrate that our proposed approach can capture additional taxonomic semantic knowledge and reduce dependence on the training dataset, outperforming the state‐of‐the‐art compared approaches on the taxonomic relation learning task.

Nowadays word embeddings, also known as word vectors, play an important role for many natural language processing (NLP) tasks. In general, these word embeddings are learned from unsupervised learning models (e.g. Word2Vec, GloVe) with a large unannotated corpus and they are independent with the task of their application. In this paper we aim to enrich word embeddings by adding more information from a specific task that is the aspect based sentiment analysis. We propose a model using a convolutional neural network that takes a labeled data set, the learned word embeddings from an unsupervised learning model (e.g. Word2Vec) as input and fine-tunes word embeddings to capture aspect category and sentiment information. We conduct experiments on restaurant review data (http://spidr-ursa.rutgers.edu/datasets/). Experimental results show that fine-tuned word embeddings outperform unsupervisedly learned word embeddings.

Distributed word embeddings have proven remarkably effective at capturing word level semantic and syntactic regularities in language for many natural language processing tasks. One recently proposed semi-supervised representation learning method called Predictive Text Embedding (PTE) utilizes both semantically labeled and unlabeled data in information networks to learn the embedding of text that produces state of-the-art performance when compared to other embedding methods. However, PTE uses supervised label information to construct one of the networks and many other possible ways of constructing such information networks are left untested. We present two unsupervised methods that can be used in constructing a large scale semantic information network from documents by combining topic models that have emerged as a powerful technique of finding useful structure in an unstructured text collection as it learns distributions over words. The first method uses Latent Dirichlet Allocation (LDA) to build a topic model over text, and constructs a word topic network with edge weights proportional to the word-topic probability distributions. The second method trains an unsupervised neural network to learn the word-document distribution, with a single hidden layer representing a topic distribution. The two weight matrices of the neural net are directly reinterpreted as the edge weights of heterogeneous text networks that can be used to train word embeddings to build an effective low dimensional representation that preserves the semantic closeness of words and documents for NLP tasks. We conduct extensive experiments to evaluate the effectiveness of our methods.

Word embeddings are essential to natural language processing tasks because they contain a single word’s syntactic and semantic information. Word embeddings have been developed widely for numerous spoken languages across the globe like English. The research community needs to pay more attention to the Urdu language despite its significant number of speakers, which amounts to approximately 231.3 million individuals. Urdu is a complex language because word boundaries in Urdu are unspecified, as it does not employ delimiters between words. The compound word, a multiword expression, is a more complex word consisting of many strings or independent base words. Traditionally, compound words are identified during the word segmentation using bigram or trigram approaches. The challenge with these techniques is that they do not produce meaningful words. This study uses morphological rule-based compound words in Urdu text documents. For text representation, a self-trained morphological rule-based compound word embedding (Cword2vec) based on the word2vec model is proposed for Urdu text sentiment analysis. The performance of self-trained morphological rule-based compound word embedding was then evaluated using four well-known deep learning models, i.e., long short-term memory (LSTM), bidirectional LSTM (BiLSTM), convolutional neural networks (CNN), and convolutional LSTM (C-LSTM) for sentiment analysis. We also compare the performance of morphological rule-based compound words with traditional compound word identification techniques such as bigrams and trigrams. Regardless of the classification model, word embedding using our proposed morphological rule-based compound words outperformed in terms of precision, recall, F1 score, and accuracy than bigrams and trigrams.

In this paper, we conduct in‐depth research and analysis on the intelligent recognition and teaching of English fuzzy text through parallel projection and region expansion. Multisense Soft Cluster Vector (MSCVec), a multisense word vector model based on nonnegative matrix decomposition and sparse soft clustering, is constructed. The MSCVec model is a monolingual word vector model, which uses nonnegative matrix decomposition of positive point mutual information between words and contexts to extract low‐rank expressions of mixed semantics of multisense words and then uses sparse. It uses the nonnegative matrix decomposition of the positive pointwise mutual information between words and contexts to extract the low‐rank expressions of the mixed semantics of the polysemous words and then uses the sparse soft clustering algorithm to partition the multiple word senses of the polysemous words and also obtains the global sense of the polysemous word affiliation distribution; the specific polysemous word cluster classes are determined based on the negative mean log‐likelihood of the global affiliation between the contextual semantics and the polysemous words, and finally, the polysemous word vectors are learned using the Fast text model under the extended dictionary word set. The advantage of the MSCVec model is that it is an unsupervised learning process without any knowledge base, and the substring representation in the model ensures the generation of unregistered word vectors; in addition, the global affiliation of the MSCVec model can also expect polysemantic word vectors to single word vectors. Compared with the traditional static word vectors, MSCVec shows excellent results in both word similarity and downstream text classification task experiments. The two sets of features are then fused and extended into new semantic features, and similarity classification experiments and stack generalization experiments are designed for comparison. In the cross‐lingual sentence‐level similarity detection task, SCLVec cross‐lingual word vector lexical‐level features outperform MSCVec multisense word vector features as the input embedding layer; deep semantic sentence‐level features trained by twin recurrent neural networks outperform the semantic features of twin convolutional neural networks; extensions of traditional statistical features can effectively improve cross‐lingual similarity detection performance, especially cross‐lingual topic model (BL‐LDA); the stack generalization integration approach maximizes the error rate of the underlying classifier and improves the detection accuracy.

Representation Learning and Learning from Limited Labeled Data for Community Question Answering

The need to capture intra-word information in natural language processing (NLP) tasks has inspired research in learning various word representations at word, character, or morpheme levels, but little attention has been given to syllables from a syllabic alphabet. Motivated by the success of compositional models in morphological languages, we present a Convolutional-long short term memory (Conv-LSTM) model for constructing Swahili word representation vectors from syllables. The unified architecture addresses the word agglutination and polysemous nature of Swahili by extracting high-level syllable features using a convolutional neural network (CNN) and then composes quality word embeddings with a long short term memory (LSTM). The word embeddings are then validated using a syllable-aware language model ( 31.267 ) and a part-of-speech (POS) tagging task ( 98.78 ), both yielding very competitive results to the state-of-art models in their respective domains. We further validate the language model using Xhosa and Shona, which are syllabic-based languages. The novelty of the study is in its capability to construct quality word embeddings from syllables using a hybrid model that does not use max-over-pool common in CNN and then the exploitation of these embeddings in POS tagging. Therefore, the study plays a crucial role in the processing of agglutinative and syllabic-based languages by contributing quality word embeddings from syllable embeddings, a robust Conv–LSTM model that learns syllables for not only language modeling and POS tagging, but also for other downstream NLP tasks.

Todays, The internet has become a source of multi-lingual content. Users are not aware of multiple languages, so the language diversity becomes a great barrier for world communication. Cross-Language Information Retrieval (CLIR) provides a solution for this language barrier, where a user can search in his native language and get the relevant information in the required language. Currently, distributed word vector representation has a trend in various Natural Language Processing (NLP) task. These word vectors are used to identify similar contextual words. In this paper, we analyze the effectiveness of word vectors across the languages in Hindi-English CLIR. Skip-Gram Model (SGM) is used to learn bi-lingual word vectors from sentence aligned comparable corpus. IBM model is used to align the source language and target language words from sentence aligned comparable corpus. Best target language translation is selected with the help of top-n word alignments and word vectors.