Various applications in the areas of computational linguistics and artificial intelligence employ semantic similarity to solve challenging tasks, such as word sense disambiguation, text classification, information retrieval, machine translation, and document clustering. Previous work on semantic similarity followed a mono-relational approach using mostly the taxonomic relation "ISA". This paper explores the benefits of using all types of non-taxonomic relations in large linked data, such as WordNet knowledge graph, to enhance existing semantic similarity and relatedness measures. We propose a holistic poly-relational approach based on a new relation-based information content and non-taxonomic-based weighted paths to devise a comprehensive semantic similarity and relatedness measure. To demonstrate the benefits of exploiting non-taxonomic relations in a knowledge graph, we used three strategies to deploy non-taxonomic relations at different granularity levels. We conducted experiments on four well-known gold standard datasets, and the results demonstrated the robustness and scalability of the proposed semantic similarity and relatedness measure, which significantly improves existing similarity measures.
Pretraining NLP models with variants of Masked Language Model (MLM) objectives has recently led to a significant improvements on many tasks. This paper examines the benefits of pretrained models as a function of the number of training samples used in the downstream task. On several text classification tasks, we show that as the number of training examples grow into the millions, the accuracy gap between finetuning BERT-based model and training vanilla LSTM from scratch narrows to within 1%. Our findings indicate that MLM-based models might reach a diminishing return point as the supervised data size increases significantly.
Exponential families are widely used in machine learning; they include many distributions in continuous and discrete domains (e.g., Gaussian, Dirichlet, Poisson, and categorical distributions via the softmax transformation). Distributions in each of these families have fixed support. In contrast, for finite domains, there has been recent work on sparse alternatives to softmax (e.g. sparsemax and alpha-entmax), which have varying support, being able to assign zero probability to irrelevant categories. This paper expands that work in two directions: first, we extend alpha-entmax to continuous domains, revealing a link with Tsallis statistics and deformed exponential families. Second, we introduce continuous-domain attention mechanisms, deriving efficient gradient backpropagation algorithms for alpha in {1,2}. Experiments on attention-based text classification, machine translation, and visual question answering illustrate the use of continuous attention in 1D and 2D, showing that it allows attending to time intervals and compact regions.
While much work on deep latent variable models of text uses continuous latent variables, discrete latent variables are interesting because they are more interpretable and typically more space efficient. We consider several approaches to learning discrete latent variable models for text in the case where exact marginalization over these variables is intractable. We compare the performance of the learned representations as features for low-resource document and sentence classification. Our best models outperform the previous best reported results with continuous representations in these low-resource settings, while learning significantly more compressed representations. Interestingly, we find that an amortized variant of Hard EM performs particularly well in the lowest-resource regimes.
The promising performance of Deep Neural Networks (DNNs) in text classification, has attracted researchers to use them for fraud review detection. However, the lack of trusted labeled data has limited the performance of the current solutions in detecting fraud reviews. The Generative Adversarial Network (GAN) as a semi-supervised method has demonstrated to be effective for data augmentation purposes. The state-of-the-art solutions utilize GANs to overcome the data scarcity problem. However, they fail to incorporate the behavioral clues in fraud generation. Additionally, state-of-the-art approaches overlook the possible bot-generated reviews in the dataset. Finally, they also suffer from a common limitation in scalability and stability of the GAN, slowing down the training procedure. In this work, we propose ScoreGAN for fraud review detection that makes use of both review text and review rating scores in the generation and detection process. Scores are incorporated through Information Gain Maximization (IGM) into the loss function for three reasons. One is to generate score-correlated reviews based on the scores given to the generator. Second, the generated reviews are employed to train the discriminator, so the discriminator can correctly label the possible bot-generated reviews through joint representations learned from the concatenation of GLobal Vector for Word representation (GLoVe) extracted from the text and the score. Finally, it can be used to improve the stability and scalability of the GAN. Results show that the proposed framework outperformed the existing state-of-the-art framework, namely FakeGAN, in terms of AP by 7\%, and 5\% on the Yelp and TripAdvisor datasets, respectively.
It is hard to train Recurrent Neural Network (RNN) with stable convergence and avoid gradient vanishing and exploding problems, as the weights in the recurrent unit are repeated from iteration to iteration. Moreover, RNN is sensitive to the initialization of weights and bias, which brings difficulties in training. The Alternating Direction Method of Multipliers (ADMM) has become a promising algorithm to train neural networks beyond traditional stochastic gradient algorithms with the gradient-free features and immunity to unsatisfactory conditions. However, ADMM could not be applied to train RNN directly since the state in the recurrent unit is repetitively updated over timesteps. Therefore, this work builds a new framework named ADMMiRNN upon the unfolded form of RNN to address the above challenges simultaneously. We also provide novel update rules and theoretical convergence analysis. We explicitly specify essential update rules in the iterations of ADMMiRNN with constructed approximation techniques and solutions to each sub-problem instead of vanilla ADMM. Numerical experiments are conducted on MNIST, IMDb, and text classification tasks. ADMMiRNN achieves convergent results and outperforms the compared baselines. Furthermore, ADMMiRNN trains RNN more stably without gradient vanishing or exploding than stochastic gradient algorithms. We also provide a distributed paralleled algorithm regarding ADMMiRNN, named P-ADMMiRNN, including Synchronous Parallel ADMMiRNN (SP-ADMMiRNN) and Asynchronous Parallel ADMMiRNN (AP-ADMMiRNN), which is the first to train RNN with ADMM in an asynchronous parallel manner. The source code is publicly available.
Machine learning algorithms have difficulties to generalize over a small set of examples. Humans can perform such a task by exploiting vast amount of background knowledge they possess. One method for enhancing learning algorithms with external knowledge is through feature generation. In this paper, we introduce a new algorithm for generating features based on a collection of auxiliary datasets. We assume that, in addition to the training set, we have access to additional datasets. Unlike the transfer learning setup, we do not assume that the auxiliary datasets represent learning tasks that are similar to our original one. The algorithm finds features that are common to the training set and the auxiliary datasets. Based on these features and examples from the auxiliary datasets, it induces predictors for new features from the auxiliary datasets. The induced predictors are then added to the original training set as generated features. Our method was tested on a variety of learning tasks, including text classification and medical prediction, and showed a significant improvement over using just the given features.
Medical concept normalization helps in discovering standard concepts in free-form text i.e., maps health-related mentions to standard concepts in a vocabulary. It is much beyond simple string matching and requires a deep semantic understanding of concept mentions. Recent research approach concept normalization as either text classification or text matching. The main drawback in existing a) text classification approaches is ignoring valuable target concepts information in learning input concept mention representation b) text matching approach is the need to separately generate target concept embeddings which is time and resource consuming. Our proposed model overcomes these drawbacks by jointly learning the representations of input concept mention and target concepts. First, it learns the input concept mention representation using RoBERTa. Second, it finds cosine similarity between embeddings of input concept mention and all the target concepts. Here, embeddings of target concepts are randomly initialized and then updated during training. Finally, the target concept with maximum cosine similarity is assigned to the input concept mention. Our model surpasses all the existing methods across three standard datasets by improving accuracy up to 2.31%.
Text classification is one of the most frequent tasks for processing textual data, facilitating among others research from large-scale datasets. Embeddings of different kinds have recently become the de facto standard as features used for text classification. These embeddings have the capacity to capture meanings of words inferred from occurrences in large external collections. While they are built out of external collections, they are unaware of the distributional characteristics of words in the classification dataset at hand, including most importantly the distribution of words across classes in training data. To make the most of these embeddings as features and to boost the performance of classifiers using them, we introduce a weighting scheme, Term Frequency-Category Ratio (TF-CR), which can weight high-frequency, category-exclusive words higher when computing word embeddings. Our experiments on eight datasets show the effectiveness of TF-CR, leading to improved performance scores over the well-known weighting schemes TF-IDF and KLD as well as over the absence of a weighting scheme in most cases.
We propose a model to tackle classification tasks in the presence of very little training data. To this aim, we approximate the notion of exact match with a theoretically sound mechanism that computes a probability of matching in the input space. Importantly, the model learns to focus on elements of the input that are relevant for the task at hand; by leveraging highlighted portions of the training data, an error boosting technique guides the learning process. In practice, it increases the error associated with relevant parts of the input by a given factor. Remarkable results on text classification tasks confirm the benefits of the proposed approach in both balanced and unbalanced cases, thus being of practical use when labeling new examples is expensive. In addition, by inspecting its weights, it is often possible to gather insights on what the model has learned.
Due to the increasing amount of data on the internet, finding a highly-informative, low-dimensional representation for text is one of the main challenges for efficient natural language processing tasks including text classification. This representation should capture the semantic information of the text while retaining their relevance level for document classification. This approach maps the documents with similar topics to a similar space in vector space representation. To obtain representation for large text, we propose the utilization of deep Siamese neural networks. To embed document relevance in topics in the distributed representation, we use a Siamese neural network to jointly learn document representations. Our Siamese network consists of two sub-network of multi-layer perceptron. We examine our representation for the text categorization task on BBC news dataset. The results show that the proposed representations outperform the conventional and state-of-the-art representations in the text classification task on this dataset.
Many text classification applications require models with satisfying performance as well as good interpretability. Traditional machine learning methods are easy to interpret but have low accuracies. The development of deep learning models boosts the performance significantly. However, deep learning models are typically hard to interpret. In this work, we propose interpretable capsule networks (iCapsNets) to bridge this gap. iCapsNets use capsules to model semantic meanings and explore novel methods to increase interpretability. The design of iCapsNets is consistent with human intuition and enables it to produce human-understandable interpretation results. Notably, iCapsNets can be interpreted both locally and globally. In terms of local interpretability, iCapsNets offer a simple yet effective method to explain the predictions for each data sample. On the other hand, iCapsNets explore a novel way to explain the model's general behavior, achieving global interpretability. Experimental studies show that our iCapsNets yield meaningful local and global interpretation results, without suffering from significant performance loss compared to non-interpretable methods.
Transfer learning, particularly approaches that combine multi-task learning with pre-trained contextualized embeddings and fine-tuning, have advanced the field of Natural Language Processing tremendously in recent years. In this paper we present MaChAmp, a toolkit for easy fine-tuning of contextualized embeddings in multi-task settings. The benefits of MaChAmp are its flexible configuration options, and the support of a variety of natural language processing tasks in a uniform toolkit, from text classification and sequence labeling to dependency parsing, masked language modeling, and text generation.
State-of-the-art NLP models can often be fooled by human-unaware transformations such as synonymous word substitution. For security reasons, it is of critical importance to develop models with certified robustness that can provably guarantee that the prediction is can not be altered by any possible synonymous word substitution. In this work, we propose a certified robust method based on a new randomized smoothing technique, which constructs a stochastic ensemble by applying random word substitutions on the input sentences, and leverage the statistical properties of the ensemble to provably certify the robustness. Our method is simple and structure-free in that it only requires the black-box queries of the model outputs, and hence can be applied to any pre-trained models (such as BERT) and any types of models (world-level or subword-level). Our method significantly outperforms recent state-of-the-art methods for certified robustness on both IMDB and Amazon text classification tasks. To the best of our knowledge, we are the first work to achieve certified robustness on large systems such as BERT with practically meaningful certified accuracy.
The BERT model has arisen as a popular state-of-the-art machine learning model in the recent years that is able to cope with multiple NLP tasks such as supervised text classification without human supervision. Its flexibility to cope with any type of corpus delivering great results has make this approach very popular not only in academia but also in the industry. Although, there are lots of different approaches that have been used throughout the years with success. In this work, we first present BERT and include a little review on classical NLP approaches. Then, we empirically test with a suite of experiments dealing different scenarios the behaviour of BERT against the traditional TF-IDF vocabulary fed to machine learning algorithms. Our purpose of this work is to add empirical evidence to support or refuse the use of BERT as a default on NLP tasks. Experiments show the superiority of BERT and its independence of features of the NLP problem such as the language of the text adding empirical evidence to use BERT as a default technique to be used in NLP problems.
Model-Agnostic Meta-Learning (MAML), a model-agnostic meta-learning method, is successfully employed in NLP applications including few-shot text classification and multi-domain low-resource language generation. Many impacting factors, including data quantity, similarity among tasks, and the balance between general language model and task-specific adaptation, can affect the performance of MAML in NLP, but few works have thoroughly studied them. In this paper, we conduct an empirical study to investigate these impacting factors and conclude when MAML works the best based on the experimental results.
We present BERTweet, the first public large-scale pre-trained language model for English Tweets. Our BERTweet, having the same architecture as BERT-base (Devlin et al., 2019), is trained using the RoBERTa pre-training procedure (Liu et al., 2019). Experiments show that BERTweet outperforms strong baselines RoBERTa-base and XLM-R-base (Conneau et al., 2020), producing better performance results than the previous state-of-the-art models on three Tweet NLP tasks: Part-of-speech tagging, Named-entity recognition and text classification. We release BERTweet under the MIT License to facilitate future research and applications on Tweet data. Our BERTweet is available at https://github.com/VinAIResearch/BERTweet
Training data for text classification is often limited in practice, especially for applications with many output classes or involving many related classification problems. This means classifiers must generalize from limited evidence, but the manner and extent of generalization is task dependent. Current practice primarily relies on pre-trained word embeddings to map words unseen in training to similar seen ones. Unfortunately, this squishes many components of meaning into highly restricted capacity. Our alternative begins with sparse pre-trained representations derived from unlabeled parsed corpora; based on the available training data, we select features that offers the relevant generalizations. This produces task-specific semantic vectors; here, we show that a feed-forward network over these vectors is especially effective in low-data scenarios, compared to existing state-of-the-art methods. By further pairing this network with a convolutional neural network, we keep this edge in low data scenarios and remain competitive when using full training sets.
The preprocessing phase is one of the key phases within the text classification pipeline. This study aims at investigating the impact of the preprocessing phase on text classification, specifically on offensive language and hate speech classification for Arabic text. The Arabic language used in social media is informal and written using Arabic dialects, which makes the text classification task very complex. Preprocessing helps in dimensionality reduction and removing useless content. We apply intensive preprocessing techniques to the dataset before processing it further and feeding it into the classification model. An intensive preprocessing-based approach demonstrates its significant impact on offensive language detection and hate speech detection shared tasks of the fourth workshop on Open-Source Arabic Corpora and Corpora Processing Tools (OSACT). Our team wins the third place (3rd) in the Sub-Task A Offensive Language Detection division and wins the first place (1st) in the Sub-Task B Hate Speech Detection division, with an F1 score of 89% and 95%, respectively, by providing the state-of-the-art performance in terms of F1, accuracy, recall, and precision for Arabic hate speech detection.
Text classification tasks which aim at harvesting and/or organizing information from electronic health records are pivotal to support clinical and translational research. However these present specific challenges compared to other classification tasks, notably due to the particular nature of the medical lexicon and language used in clinical records. Recent advances in embedding methods have shown promising results for several clinical tasks, yet there is no exhaustive comparison of such approaches with other commonly used word representations and classification models. In this work, we analyse the impact of various word representations, text pre-processing and classification algorithms on the performance of four different text classification tasks. The results show that traditional approaches, when tailored to the specific language and structure of the text inherent to the classification task, can achieve or exceed the performance of more recent ones based on contextual embeddings such as BERT.