Adversarial attack aims to perturb input sequences and mislead a trained model for false predictions. To enhance the model robustness, defensing methods are accordingly employed by either data augmentation (involving adversarial samples) or model enhancement (modifying the training loss and/or model architecture). In contrast to previous work, this paper revisits the masked language modeling (MLM) and presents a simple yet efficient algorithm against adversarial attacks, termed [MASK] insertion for defensing (MI4D). Specifically, MI4D simply inserts [MASK] tokens to input sequences during training and inference, maximizing the intersection of the new convex hull (MI4D creates) with the original one (the clean input forms). As neither additional adversarial samples nor the model modification is required, MI4D is as computationally efficient as traditional fine-tuning. Comprehensive experiments have been conducted using three benchmark datasets and four attacking methods. MI4D yields a significant improvement (on average) of the accuracy between 3.2 and 11.1 absolute points when compared with six state-of-the-art defensing baselines.

We introduce YATO, an open-source, easy-to-use toolkit for text analysis with deep learning. Different from existing heavily engineered toolkits and platforms, YATO is lightweight and user-friendly for researchers from cross-disciplinary areas. Designed in a hierarchical structure, YATO supports free combinations of three types of widely used features including 1) traditional neural networks (CNN, RNN, etc.); 2) pre-trained language models (BERT, RoBERTa, ELECTRA, etc.); and 3) user-customized neural features via a simple configurable file. Benefiting from the advantages of flexibility and ease of use, YATO can facilitate fast reproduction and refinement of state-of-the-art NLP models, and promote the cross-disciplinary applications of NLP techniques. The code, examples, and documentation are publicly available at https://github.com/jiesutd/YATO. A demo video is also available at https://www.youtube.com/playlist?list=PLJ0mhzMcRuDUlTkzBfAftOqiJRxYTTjXH.

Quality control is essential for creating extractive question answering (EQA) datasets via crowdsourcing. Aggregation across answers, i.e. word spans within passages annotated, by different crowd workers is one major focus for ensuring its quality. However, crowd workers cannot reach a consensus on a considerable portion of questions. We introduce a simple yet effective answer aggregation method that takes into account the relations among the answer, question, and context passage. We evaluate answer quality from both the view of question answering model to determine how confident the QA model is about each answer and the view of the answer verification model to determine whether the answer is correct. Then we compute aggregation scores with each answer’s quality and its contextual embedding produced by pre-trained language models. The experiments on a large real crowdsourced EQA dataset show that our framework outperforms baselines by around 16% on precision and effectively conduct answer aggregation for extractive QA task.

Drug-drug interaction (DDI) may leads to adverse reactions in patients, thus it is important to extract such knowledge from biomedical texts. However, previously proposed approaches typically focus on capturing sentence-aspect information while ignoring valuable knowledge concerning the whole corpus. In this paper, we propose a Multi-aspect Graph-based DDI extraction model, named DDI-MuG. We first employ a bio-specific pre-trained language model to obtain the token contextualized representations. Then we use two graphs to get syntactic information from input instance and word co-occurrence information within the entire corpus, respectively. Finally, we combine the representations of drug entities and verb tokens for the final classification. It is encouraging to see that the proposed model outperforms all baseline models on two benchmark datasets. To the best of our knowledge, this is the first model that explores multi-aspect graphs to the DDI extraction task, and we hope it can establish a foundation for more robust multi-aspect works in the future.

Current news datasets merely focus on text features on the news and rarely leverage the feature of images, excluding numerous essential features for news classification. In this paper, we propose a new dataset, N24News, which is generated from New York Times with 24 categories and contains both text and image information in each news. We use a multitask multimodal method and the experimental results show multimodal news classification performs better than text-only news classification. Depending on the length of the text, the classification accuracy can be increased by up to 8.11%. Our research reveals the relationship between the performance of a multimodal classifier and its sub-classifiers, and also the possible improvements when applying multimodal in news classification. N24News is shown to have great potential to prompt the multimodal news studies.

We investigate subword information for Chinese word segmentation, by integrating sub word embeddings trained using byte-pair encoding into a Lattice LSTM (LaLSTM) network over a character sequence. Experiments on standard benchmark show that subword information brings significant gains over strong character-based segmentation models. To our knowledge, this is the first research on the effectiveness of subwords on neural word segmentation.

Social media provides a timely yet challenging data source for adverse drug reaction (ADR) detection. Existing dictionary-based, semi-supervised learning approaches are intrinsically limited by the coverage and maintainability of laymen health vocabularies. In this paper, we introduce a data augmentation approach that leverages variational autoencoders to learn high-quality data distributions from a large unlabeled dataset, and subsequently, to automatically generate a large labeled training set from a small set of labeled samples. This allows for efficient social-media ADR detection with low training and re-training costs to adapt to the changes and emergence of informal medical laymen terms. An extensive evaluation performed on Twitter and Reddit data shows that our approach matches the performance of fully-supervised approaches while requiring only 25% of training data.

We investigate the design challenges of constructing effective and efficient neural sequence labeling systems, by reproducing twelve neural sequence labeling models, which include most of the state-of-the-art structures, and conduct a systematic model comparison on three benchmarks (i.e. NER, Chunking, and POS tagging). Misconceptions and inconsistent conclusions in existing literature are examined and clarified under statistical experiments. In the comparison and analysis process, we reach several practical conclusions which can be useful to practitioners.

We investigate a lattice-structured LSTM model for Chinese NER, which encodes a sequence of input characters as well as all potential words that match a lexicon. Compared with character-based methods, our model explicitly leverages word and word sequence information. Compared with word-based methods, lattice LSTM does not suffer from segmentation errors. Gated recurrent cells allow our model to choose the most relevant characters and words from a sentence for better NER results. Experiments on various datasets show that lattice LSTM outperforms both word-based and character-based LSTM baselines, achieving the best results.

In this paper, we introduce Yedda, a lightweight but efficient and comprehensive open-source tool for text span annotation. Yedda provides a systematic solution for text span annotation, ranging from collaborative user annotation to administrator evaluation and analysis. It overcomes the low efficiency of traditional text annotation tools by annotating entities through both command line and shortcut keys, which are configurable with custom labels. Yedda also gives intelligent recommendations by learning the up-to-date annotated text. An administrator client is developed to evaluate annotation quality of multiple annotators and generate detailed comparison report for each annotator pair. Experiments show that the proposed system can reduce the annotation time by half compared with existing annotation tools. And the annotation time can be further compressed by 16.47% through intelligent recommendation.

This paper describes NCRF++, a toolkit for neural sequence labeling. NCRF++ is designed for quick implementation of different neural sequence labeling models with a CRF inference layer. It provides users with an inference for building the custom model structure through configuration file with flexible neural feature design and utilization. Built on PyTorch http://pytorch.org/, the core operations are calculated in batch, making the toolkit efficient with the acceleration of GPU. It also includes the implementations of most state-of-the-art neural sequence labeling models such as LSTM-CRF, facilitating reproducing and refinement on those methods.

Neural word segmentation research has benefited from large-scale raw texts by leveraging them for pretraining character and word embeddings. On the other hand, statistical segmentation research has exploited richer sources of external information, such as punctuation, automatic segmentation and POS. We investigate the effectiveness of a range of external training sources for neural word segmentation by building a modular segmentation model, pretraining the most important submodule using rich external sources. Results show that such pretraining significantly improves the model, leading to accuracies competitive to the best methods on six benchmarks.

Singlish can be interesting to the ACL community both linguistically as a major creole based on English, and computationally for information extraction and sentiment analysis of regional social media. We investigate dependency parsing of Singlish by constructing a dependency treebank under the Universal Dependencies scheme, and then training a neural network model by integrating English syntactic knowledge into a state-of-the-art parser trained on the Singlish treebank. Results show that English knowledge can lead to 25% relative error reduction, resulting in a parser of 84.47% accuracies. To the best of our knowledge, we are the first to use neural stacking to improve cross-lingual dependency parsing on low-resource languages. We make both our annotation and parser available for further research.

Neural network models have recently been applied to the task of automatic essay scoring, giving promising results. Existing work used recurrent neural networks and convolutional neural networks to model input essays, giving grades based on a single vector representation of the essay. On the other hand, the relative advantages of RNNs and CNNs have not been compared. In addition, different parts of the essay can contribute differently for scoring, which is not captured by existing models. We address these issues by building a hierarchical sentence-document model to represent essays, using the attention mechanism to automatically decide the relative weights of words and sentences. Results show that our model outperforms the previous state-of-the-art methods, demonstrating the effectiveness of the attention mechanism.

We propose a neural reranking system for named entity recognition (NER), leverages recurrent neural network models to learn sentence-level patterns that involve named entity mentions. In particular, given an output sentence produced by a baseline NER model, we replace all entity mentions, such as Barack Obama, into their entity types, such as PER. The resulting sentence patterns contain direct output information, yet is less sparse without specific named entities. For example, “PER was born in LOC” can be such a pattern. LSTM and CNN structures are utilised for learning deep representations of such sentences for reranking. Results show that our system can significantly improve the NER accuracies over two different baselines, giving the best reported results on a standard benchmark.

We present a light-weight machine learning tool for NLP research. The package supports operations on both discrete and dense vectors, facilitating implementation of linear models as well as neural models. It provides several basic layers which mainly aims for single-layer linear and non-linear transformations. By using these layers, we can conveniently implement linear models and simple neural models. Besides, this package also integrates several complex layers by composing those basic layers, such as RNN, Attention Pooling, LSTM and gated RNN. Those complex layers can be used to implement deep neural models directly.

Recent improvements in speech recognition technology have resulted in products that can now demonstrate commercial value in a variety of applications. Many vendors are marketing products which combine ASR applications including continuous dictation, command-and-control interfaces, and transcription of recorded speech at an accuracy of 98%. In this study, we measured the accuracy of certain commercially available desktop speech recognition engines in multiple languages. Using word error rate as a benchmark, this work compares recognition accuracy across eight languages and the products of three manufacturers. Results show that two systems performed almost the same while a third system recognized at lower accuracy, although none of the systems reached the claimed accuracy. Read speech was recognized better than spontaneous speech. The systems for US-English, Japanese and Spanish showed higher accuracy than the systems for UK-English, German, French and Chinese.