Using the shared-private paradigm and adversarial training has significantly improved the performances of multi-domain text classification (MDTC) models. However, there are two issues for the existing methods. First, instances from the multiple domains are not sufficient for domain-invariant feature extraction. Second, aligning on the marginal distributions may lead to fatal mismatching. In this paper, we propose a mixup regularized adversarial network (MRAN) to address these two issues. More specifically, the domain and category mixup regularizations are introduced to enrich the intrinsic features in the shared latent space and enforce consistent predictions in-between training instances such that the learned features can be more domain-invariant and discriminative. We conduct experiments on two benchmarks: The Amazon review dataset and the FDU-MTL dataset. Our approach on these two datasets yields average accuracies of 87.64\% and 89.0\% respectively, outperforming all relevant baselines.
Text classification is the most basic natural language processing task. It has a wide range of applications ranging from sentiment analysis to topic classification. Recently, deep learning approaches based on CNN, LSTM, and Transformers have been the de facto approach for text classification. In this work, we highlight a common issue associated with these approaches. We show that these systems are over-reliant on the important words present in the text that are useful for classification. With limited training data and discriminative training strategy, these approaches tend to ignore the semantic meaning of the sentence and rather just focus on keywords or important n-grams. We propose a simple black box technique ShutText to present the shortcomings of the model and identify the over-reliance of the model on keywords. This involves randomly shuffling the words in a sentence and evaluating the classification accuracy. We see that on common text classification datasets there is very little effect of shuffling and with high probability these models predict the original class. We also evaluate the effect of language model pretraining on these models and try to answer questions around model robustness to out of domain sentences. We show that simple models based on CNN or LSTM as well as complex models like BERT are questionable in terms of their syntactic and semantic understanding.
Federated learning (FL) is a promising approach to distributed compute, as well as distributed data, and provides a level of privacy and compliance to legal frameworks. This makes FL attractive for both consumer and healthcare applications. While the area is actively being explored, few studies have examined FL in the context of larger language models and there is a lack of comprehensive reviews of robustness across tasks, architectures, numbers of clients, and other relevant factors. In this paper, we explore the fine-tuning of Transformer-based language models in a federated learning setting. We evaluate three popular BERT-variants of different sizes (BERT, ALBERT, and DistilBERT) on a number of text classification tasks such as sentiment analysis and author identification. We perform an extensive sweep over the number of clients, ranging up to 32, to evaluate the impact of distributed compute on task performance in the federated averaging setting. While our findings suggest that the large sizes of the evaluated models are not generally prohibitive to federated training, we found that the different models handle federated averaging to a varying degree. Most notably, DistilBERT converges significantly slower with larger numbers of clients, and under some circumstances, even collapses to chance level performance. Investigating this issue presents an interesting perspective for future research.
In this work, we propose the use of a fully managed machine learning service, which utilizes active learning to directly build models from unstructured data. With this tool, business users can quickly and easily build machine learning models and then directly deploy them into a production ready hosted environment without much involvement from data scientists. Our approach leverages state-of-the-art text representation like OpenAI's GPT2 and a fast implementation of the active learning workflow that relies on a simple construction of incremental learning using linear models, thus providing a brisk and efficient labeling experience for the users. Experiments on both publicly available and real-life insurance datasets empirically show why our choices of simple and fast classification algorithms are ideal for the task at hand.
Modern classification models tend to struggle when the amount of annotated data is scarce. To overcome this issue, several neural few-shot classification models have emerged, yielding significant progress over time, both in Computer Vision and Natural Language Processing. In the latter, such models used to rely on fixed word embeddings before the advent of transformers. Additionally, some models used in Computer Vision are yet to be tested in NLP applications. In this paper, we compare all these models, first adapting those made in the field of image processing to NLP, and second providing them access to transformers. We then test these models equipped with the same transformer-based encoder on the intent detection task, known for having a large number of classes. Our results reveal that while methods perform almost equally on the ARSC dataset, this is not the case for the Intent Detection task, where the most recent and supposedly best competitors perform worse than older and simpler ones (while all are given access to transformers). We also show that a simple baseline is surprisingly strong. All the new developed models, as well as the evaluation framework, are made publicly available.
In the area of geographic information processing. There are few researches on geographic text classification. However, the application of this task in Chinese is relatively rare. In our work, we intend to implement a method to extract text containing geographical entities from a large number of network text. The geographic information in these texts is of great practical significance to transportation, urban and rural planning, disaster relief and other fields. We use the method of graph convolutional neural network with attention mechanism to achieve this function. Graph attention networks is an improvement of graph convolutional neural networks. Compared with GCN, the advantage of GAT is that the attention mechanism is proposed to weight the sum of the characteristics of adjacent nodes. In addition, We construct a Chinese dataset containing geographical classification from multiple datasets of Chinese text classification. The Macro-F Score of the geoGAT we used reached 95\% on the new Chinese dataset.
Large datasets in NLP suffer from noisy labels, due to erroneous automatic and human annotation procedures. We study the problem of text classification with label noise, and aim to capture this noise through an auxiliary noise model over the classifier. We first assign a probability score to each training sample of having a noisy label, through a beta mixture model fitted on the losses at an early epoch of training. Then, we use this score to selectively guide the learning of the noise model and classifier. Our empirical evaluation on two text classification tasks shows that our approach can improve over the baseline accuracy, and prevent over-fitting to the noise.
Multi-label text classification (MLTC) aims to annotate documents with the most relevant labels from a number of candidate labels. In real applications, the distribution of label frequency often exhibits a long tail, i.e., a few labels are associated with a large number of documents (a.k.a. head labels), while a large fraction of labels are associated with a small number of documents (a.k.a. tail labels). To address the challenge of insufficient training data on tail label classification, we propose a Head-to-Tail Network (HTTN) to transfer the meta-knowledge from the data-rich head labels to data-poor tail labels. The meta-knowledge is the mapping from few-shot network parameters to many-shot network parameters, which aims to promote the generalizability of tail classifiers. Extensive experimental results on three benchmark datasets demonstrate that HTTN consistently outperforms the state-of-the-art methods. The code and hyper-parameter settings are released for reproducibility
Due to biases introduced by large real-world datasets, deviations of deep learning models from their expected behavior on out-of-distribution test data are worrisome. Especially when data come from imbalanced or heavy-tailed label distributions or minority groups of a sensitive feature. Classical approaches to address these biases are mostly data- or application-dependent, hence are burdensome to tune. Some meta-learning approaches, on the other hand, aim to learn hyperparameters in the learning process using different objective functions on training and validation data. However, these methods suffer from high computational complexity and are not scalable to large datasets. In this paper, we propose a unified data-driven regularization approach to learn a generalizable model from biased data. The proposed framework, named targeted data-driven regularization (TDR), is model- and dataset-agnostic and employs a target dataset that resembles the desired nature of test data in order to guide the learning process in a coupled manner. We cast the problem as a bilevel optimization and propose an efficient stochastic gradient descent based method to solve it. The framework can be utilized to alleviate various types of biases in real-world applications. We empirically show, on both synthetic and real-world datasets, the superior performance of TDR for resolving issues stem from these biases.
In this paper, we present a transfer learning system to perform technical domain identification on multilingual text data. We have submitted two runs, one uses the transformer model BERT, and the other uses XLM-ROBERTa with the CNN model for text classification. These models allowed us to identify the domain of the given sentences for the ICON 2020 shared Task, TechDOfication: Technical Domain Identification. Our system ranked the best for the subtasks 1d, 1g for the given TechDOfication dataset.
Text classification is a fundamental problem in the field of natural language processing. Text classification mainly focuses on giving more importance to all the relevant features that help classify the textual data. Apart from these, the text can have redundant or highly correlated features. These features increase the complexity of the classification algorithm. Thus, many dimensionality reduction methods were proposed with the traditional machine learning classifiers. The use of dimensionality reduction methods with machine learning classifiers has achieved good results. In this paper, we propose a hybrid feature selection method for obtaining relevant features by combining various filter-based feature selection methods and fastText classifier. We then present three ways of implementing a feature selection and neural network pipeline. We observed a reduction in training time when feature selection methods are used along with neural networks. We also observed a slight increase in accuracy on some datasets.
We describe latent Dirichlet allocation (LDA), a generative probabilistic model for collections of discrete data such as text corpora. LDA is a three-level hierarchical Bayesian model, in which each item of a collection is modeled as a finite mixture over an underlying set of topics. Each topic is, in turn, modeled as an infinite mixture over an underlying set of topic probabilities. In the context of text modeling, the topic probabilities provide an explicit representation of a document. We present efficient approximate inference techniques based on variational methods and an EM algorithm for empirical Bayes parameter estimation. We report results in document modeling, text classification, and collaborative filtering, comparing to a mixture of unigrams model and the probabilistic LSI model.
Deep learning models are susceptible to adversarial examples that have imperceptible perturbations in the original input, resulting in adversarial attacks against these models. Analysis of these attacks on the state of the art transformers in NLP can help improve the robustness of these models against such adversarial inputs. In this paper, we present Adv-OLM, a black-box attack method that adapts the idea of Occlusion and Language Models (OLM) to the current state of the art attack methods. OLM is used to rank words of a sentence, which are later substituted using word replacement strategies. We experimentally show that our approach outperforms other attack methods for several text classification tasks.
We investigate the self-attention mechanism of BERT in a fine-tuning scenario for the classification of scientific articles over a taxonomy of research disciplines. We observe how self-attention focuses on words that are highly related to the domain of the article. Particularly, a small subset of vocabulary words tends to receive most of the attention. We compare and evaluate the subset of the most attended words with feature selection methods normally used for text classification in order to characterize self-attention as a possible feature selection approach. Using ConceptNet as ground truth, we also find that attended words are more related to the research fields of the articles. However, conventional feature selection methods are still a better option to learn classifiers from scratch. This result suggests that, while self-attention identifies domain-relevant terms, the discriminatory information in BERT is encoded in the contextualized outputs and the classification layer. It also raises the question whether injecting feature selection methods in the self-attention mechanism could further optimize single sequence classification using transformers.
We propose a privacy-preserving Naive Bayes classifier and apply it to the problem of private text classification. In this setting, a party (Alice) holds a text message, while another party (Bob) holds a classifier. At the end of the protocol, Alice will only learn the result of the classifier applied to her text input and Bob learns nothing. Our solution is based on Secure Multiparty Computation (SMC). Our Rust implementation provides a fast and secure solution for the classification of unstructured text. Applying our solution to the case of spam detection (the solution is generic, and can be used in any other scenario in which the Naive Bayes classifier can be employed), we can classify an SMS as spam or ham in less than 340ms in the case where the dictionary size of Bob's model includes all words (n = 5200) and Alice's SMS has at most m = 160 unigrams. In the case with n = 369 and m = 8 (the average of a spam SMS in the database), our solution takes only 21ms.
The advent of the Internet and a large number of digital technologies has brought with it many different challenges. A large amount of data is found on the web, which in most cases is unstructured and unorganized, and this contributes to the fact that the use and manipulation of this data is quite a difficult process. Due to this fact, the usage of different machine and deep learning techniques for Text Classification has gained its importance, which improved this discipline and made it more interesting for scientists and researchers for further study. This paper aims to classify the pedagogical content using two different models, the K-Nearest Neighbor (KNN) from the conventional models and the Long short-term memory (LSTM) recurrent neural network from the deep learning models. The result indicates that the accuracy of classifying the pedagogical content reaches 92.52 % using KNN model and 87.71 % using LSTM model.
Recent improvements in the predictive quality of natural language processing systems are often dependent on a substantial increase in the number of model parameters. This has led to various attempts of compressing such models, but existing methods have not considered the differences in the predictive power of various model components or in the generalizability of the compressed models. To understand the connection between model compression and out-of-distribution generalization, we define the task of compressing language representation models such that they perform best in a domain adaptation setting. We choose to address this problem from a causal perspective, attempting to estimate the average treatment effect (ATE) of a model component, such as a single layer, on the model's predictions. Our proposed ATE-guided Model Compression scheme (AMoC), generates many model candidates, differing by the model components that were removed. Then, we select the best candidate through a stepwise regression model that utilizes the ATE to predict the expected performance on the target domain. AMoC outperforms strong baselines on dozens of domain pairs across three text classification and sequence tagging tasks.
Recent advancements in language models based on recurrent neural networks and transformers architecture have achieved state-of-the-art results on a wide range of natural language processing tasks such as pos tagging, named entity recognition, and text classification. However, most of these language models are pre-trained in high resource languages like English, German, Spanish. Multi-lingual language models include Indian languages like Hindi, Telugu, Bengali in their training corpus, but they often fail to represent the linguistic features of these languages as they are not the primary language of the study. We introduce HinFlair, which is a language representation model (contextual string embeddings) pre-trained on a large monolingual Hindi corpus. Experiments were conducted on 6 text classification datasets and a Hindi dependency treebank to analyze the performance of these contextualized string embeddings for the Hindi language. Results show that HinFlair outperforms previous state-of-the-art publicly available pre-trained embeddings for downstream tasks like text classification and pos tagging. Also, HinFlair when combined with FastText embeddings outperforms many transformers-based language models trained particularly for the Hindi language.
On-device Deep Neural Networks (DNNs) have recently gained more attention due to the increasing computing power of the mobile devices and the number of applications in Computer Vision (CV), Natural Language Processing (NLP), and Internet of Things (IoTs). Unfortunately, the existing efficient convolutional neural network (CNN) architectures designed for CV tasks are not directly applicable to NLP tasks and the tiny Recurrent Neural Network (RNN) architectures have been designed primarily for IoT applications. In NLP applications, although model compression has seen initial success in on-device text classification, there are at least three major challenges yet to be addressed: adversarial robustness, explainability, and personalization. Here we attempt to tackle these challenges by designing a new training scheme for model compression and adversarial robustness, including the optimization of an explainable feature mapping objective, a knowledge distillation objective, and an adversarially robustness objective. The resulting compressed model is personalized using on-device private training data via fine-tuning. We perform extensive experiments to compare our approach with both compact RNN (e.g., FastGRNN) and compressed RNN (e.g., PRADO) architectures in both natural and adversarial NLP test settings.
Traditional data augmentation aims to increase the coverage of the input distribution by generating augmented examples that strongly resemble original samples in an online fashion where augmented examples dominate training. In this paper, we propose an alternative perspective -- a multi-task view (MTV) of data augmentation -- in which the primary task trains on original examples and the auxiliary task trains on augmented examples. In MTV data augmentation, both original and augmented samples are weighted substantively during training, relaxing the constraint that augmented examples must resemble original data and thereby allowing us to apply stronger levels of augmentation. In empirical experiments using four common data augmentation techniques on three benchmark text classification datasets, we find that the MTV leads to higher and more robust performance improvements than traditional augmentation.