COLING 2020 论文列表

Causality represents the most important kind of correlation between events. Extracting causali-ty from text has become a promising hot topic in NLP. However, there is no mature research systems and datasets for public evaluation. Moreover, there is a lack of unified causal sequence label methods, which constitute the key factors that hinder the progress of causality extraction research. We survey the limitations and shortcomings of existing causality research field com-prehensively from the aspects of basic concepts, extraction methods, experimental data, and la-bel methods, so as to provide reference for future research on causality extraction. We summa-rize the existing causality datasets, explore their practicability and extensibility from multiple perspectives and create a new causal dataset ESC. Aiming at the problem of causal sequence labeling, we analyse the existing methods with a summarization of its regulation and propose a new causal label method of core word. Multiple candidate causal label sequences are put for-ward according to label controversy to explore the optimal label method through experiments, and suggestions are provided for selecting label method.

We present a novel retrofitting model that can leverage relational knowledge available in a knowledge resource to improve word embeddings. The knowledge is captured in terms of relation inequality constraints that compare similarity of related and unrelated entities in the context of an anchor entity. These constraints are used as training data to learn a non-linear transformation function that maps original word vectors to a vector space respecting these constraints. The transformation function is learned in a similarity metric learning setting using Triplet network architecture. We applied our model to synonymy, antonymy and hypernymy relations in WordNet and observed large gains in performance over original distributional models as well as other retrofitting approaches on word similarity task and significant overall improvement on lexical entailment detection task.

Generating knowledge from natural language data has aided in solving many artificial intelligence problems. Vector representations of words have been the driving force behind the majority of natural language processing tasks. This paper develops a novel approach for predicting the conservation status of animal species using custom generated scientific name embeddings. We use two different vector embeddings generated using representation learning on Wikipedia text and animal taxonomy data. We generate name embeddings for all species in the animal kingdom using unsupervised learning and build a model on the IUCN Red List dataset to classify species into endangered or least-concern. To our knowledge, this is the first work that makes use of learnt features instead of handcrafted features for this task and achieves competitive results. Based on the high confidence results of our model, we also predict the conservation status of data deficient species whose conservation status is still unknown and thus steering more focus towards them for protection. These embeddings have also been made publicly available here. We believe this will greatly help in solving various downstream tasks and further advance research in the cross-domain involving natural language processing, conservation biology, and life sciences.

Coreference resolution is the task of identifying all mentions in a text that refer to the same real-world entity. Collecting sufficient labelled data from expert annotators to train a high-performance coreference resolution system is time-consuming and expensive. Crowdsourcing makes it possible to obtain the required amounts of data rapidly and cost-effectively. However, crowd-sourced labels can be noisy. To ensure high-quality data, it is crucial to infer the correct labels by aggregating the noisy labels. In this paper, we split the aggregation into two subtasks, i.e, mention classification and coreference chain inference. Firstly, we predict the general class of each mention using an autoencoder, which incorporates contextual information about each mention, while at the same time taking into account the mention’s annotation complexity and annotators’ reliability at different levels. Secondly, to determine the coreference chain of each mention, we use weighted voting which takes into account the learned reliability in the first subtask. Experimental results demonstrate the effectiveness of our method in predicting the correct labels. We also illustrate our model’s interpretability through a comprehensive analysis of experimental results.

Local coherence relation between two phrases/sentences such as cause-effect and contrast gives a strong influence of whether a text is well-structured or not. This paper follows the assumption and presents a method for scoring text clarity by utilizing local coherence between adjacent sentences. We hypothesize that the contextual features of coherence relations learned by utilizing different data from the target training data are also possible to discriminate well-structured of the target text and thus help to score the text clarity. We propose a text clarity scoring method that utilizes local coherence analysis with an out-domain setting, i.e. the training data for the source and target tasks are different from each other. The method with language model pre-training BERT firstly trains the local coherence model as an auxiliary manner and then re-trains it together with clarity text scoring model. The experimental results by using the PeerRead benchmark dataset show the improvement compared with a single model, scoring text clarity model. Our source codes are available online.

Natural language understanding (NLU) aims at identifying user intent and extracting semantic slots. This requires sufficient annotating data to get considerable performance in real-world situations. Active learning (AL) has been well-studied to decrease the needed amount of the annotating data and successfully applied to NLU. However, no research has been done on investigating how the relation information between intents and slots can improve the efficiency of AL algorithms. In this paper, we propose a multitask AL framework for NLU. Our framework enables pool-based AL algorithms to make use of the relation information between sub-tasks provided by a joint model, and we propose an efficient computation for the entropy of a joint model. Experimental results show our framework can achieve competitive performance with less training data than baseline methods on all datasets. We also demonstrate that when using the entropy as the query strategy, the model with complete relation information can perform better than those with partial information. Additionally, we demonstrate that the efficiency of these active learning algorithms in our framework is still effective when incorporate with the Bidirectional Encoder Representations from Transformers (BERT).

This paper presents M-GAAMA, a Multilingual Question Answering architecture and demo system. This is the first multilingual machine reading comprehension (MRC) demo which is able to answer questions in over 100 languages. M-GAAMA answers questions from a given passage in the same or different language. It incorporates several existing multilingual models that can be used interchangeably in the demo such as M-BERT and XLM-R. The M-GAAMA demo also improves language accessibility by incorporating the IBM Watson machine translation widget to provide additional capabilities to the user to see an answer in their desired language. We also show how M-GAAMA can be used in downstream tasks by incorporating it into an END-TO-END-QA system using CFO (Chakravarti et al., 2019). We experiment with our system architecture on the Multi-Lingual Question Answering (MLQA) and the COVID-19 CORD (Wang et al., 2020; Tang et al., 2020) datasets to provide insights into the performance of the system.

Various methods have already been proposed for learning entity embeddings from text descriptions. Such embeddings are commonly used for inferring properties of entities, for recommendation and entity-oriented search, and for injecting background knowledge into neural architectures, among others. Entity embeddings essentially serve as a compact encoding of a similarity relation, but similarity is an inherently multi-faceted notion. By representing entities as single vectors, existing methods leave it to downstream applications to identify these different facets, and to select the most relevant ones. In this paper, we propose a model that instead learns several vectors for each entity, each of which intuitively captures a different aspect of the considered domain. We use a mixture-of-experts formulation to jointly learn these facet-specific embeddings. The individual entity embeddings are learned using a variant of the GloVe model, which has the advantage that we can easily identify which properties are modelled well in which of the learned embeddings. This is exploited by an associated gating network, which uses pre-trained word vectors to encourage the properties that are modelled by a given embedding to be semantically coherent, i.e. to encourage each of the individual embeddings to capture a meaningful facet.

Learning a mapping between word embeddings of two languages given a dictionary is an important problem with several applications. A common mapping approach is using an orthogonal matrix. The Orthogonal Procrustes Analysis (PA) algorithm can be applied to find the optimal orthogonal matrix. This solution restricts the expressiveness of the translation model which may result in sub-optimal translations. We propose a natural extension of the PA algorithm that uses multiple orthogonal translation matrices to model the mapping and derive an algorithm to learn these multiple matrices. We achieve better performance in a bilingual word translation task and a cross-lingual word similarity task compared to the single matrix baseline. We also show how multiple matrices can model multiple senses of a word.

This paper reports on a structured evaluation of feature-based Machine Learning algorithms for selecting the form of a referring expression in discourse context. Based on this evaluation, we selected seven feature sets from the literature, amounting to 65 distinct linguistic features. The features were then grouped into 9 broad classes. After building Random Forest models, we used Feature Importance Ranking and Sequential Forward Search methods to assess the “importance” of the features. Combining the results of the two methods, we propose a consensus feature set. The 6 features in our consensus set come from 4 different classes, namely grammatical role, inherent features of the referent, antecedent form and recency.

Paraphrase generation (PG) is of great importance to many downstream tasks in natural language processing. Diversity is an essential nature to PG for enhancing generalization capability and robustness of downstream applications. Recently, neural sequence-to-sequence (Seq2Seq) models have shown promising results in PG. However, traditional model training for PG focuses on optimizing model prediction against single reference and employs cross-entropy loss, which objective is unable to encourage model to generate diverse paraphrases. In this work, we present a novel approach with multi-objective learning to PG. We propose a learning-exploring method to generate sentences as learning objectives from the learned data distribution, and employ reinforcement learning to combine these new learning objectives for model training. We first design a sample-based algorithm to explore diverse sentences. Then we introduce several reward functions to evaluate the sampled sentences as learning signals in terms of expressive diversity and semantic fidelity, aiming to generate diverse and high-quality paraphrases. To effectively optimize model performance satisfying different evaluating aspects, we use a GradNorm-based algorithm that automatically balances these training objectives. Experiments and analyses on Quora and Twitter datasets demonstrate that our proposed method not only gains a significant increase in diversity but also improves generation quality over several state-of-the-art baselines.

We present a large-scale corpus of e-mail conversations with domain-agnostic and two-level dialogue act (DA) annotations towards the goal of a better understanding of asynchronous conversations. We annotate over 6,000 messages and 35,000 sentences from more than 2,000 threads. For a domain-independent and application-independent DA annotations, we choose ISO standard 24617-2 as the annotation scheme. To assess the difficulty of DA recognition on our corpus, we evaluate several models, including a pre-trained contextual representation model, as our baselines. The experimental results show that BERT outperforms other neural network models, including previous state-of-the-art models, but falls short of a human performance. We also demonstrate that DA tags of two-level granularity enable a DA recognition model to learn efficiently by using multi-task learning. An evaluation of a model trained on our corpus against other domains of asynchronous conversation reveals the domain independence of our DA annotations.

Keyphrase extraction is the task of extracting a small set of phrases that best describe a document. Most existing benchmark datasets for the task typically have limited numbers of annotated documents, making it challenging to train increasingly complex neural networks. In contrast, digital libraries store millions of scientific articles online, covering a wide range of topics. While a significant portion of these articles contain keyphrases provided by their authors, most other articles lack such kind of annotations. Therefore, to effectively utilize these large amounts of unlabeled articles, we propose a simple and efficient joint learning approach based on the idea of self-distillation. Experimental results show that our approach consistently improves the performance of baseline models for keyphrase extraction. Furthermore, our best models outperform previous methods for the task, achieving new state-of-the-art results on two public benchmarks: Inspec and SemEval-2017.

Most current state-of-the art systems for generating English text from Abstract Meaning Representation (AMR) have been evaluated only using automated metrics, such as BLEU, which are known to be problematic for natural language generation. In this work, we present the results of a new human evaluation which collects fluency and adequacy scores, as well as categorization of error types, for several recent AMR generation systems. We discuss the relative quality of these systems and how our results compare to those of automatic metrics, finding that while the metrics are mostly successful in ranking systems overall, collecting human judgments allows for more nuanced comparisons. We also analyze common errors made by these systems.

Motivated by applications such as question answering, fact checking, and data integration, there is significant interest in constructing knowledge graphs by extracting information from unstructured information sources, particularly text documents. Knowledge graphs have emerged as a standard for structured knowledge representation, whereby entities and their inter-relations are represented and conveniently stored as (subject,predicate,object) triples in a graph that can be used to power various downstream applications. The proliferation of financial news sources reporting on companies, markets, currencies, and stocks presents an opportunity for extracting valuable knowledge about this crucial domain. In this paper, we focus on constructing a knowledge graph automatically by information extraction from a large corpus of financial news articles. For that purpose, we develop a high precision knowledge extraction pipeline tailored for the financial domain. This pipeline combines multiple information extraction techniques with a financial dictionary that we built, all working together to produce over 342,000 compact extractions from over 288,000 financial news articles, with a precision of 78% at the top-100 extractions. The extracted triples are stored in a knowledge graph making them readily available for use in downstream applications.

The abundant semi-structured data on the Web, such as HTML-based tables and lists, provide commercial search engines a rich information source for question answering (QA). Different from plain text passages in Web documents, Web tables and lists have inherent structures, which carry semantic correlations among various elements in tables and lists. Many existing studies treat tables and lists as flat documents with pieces of text and do not make good use of semantic information hidden in structures. In this paper, we propose a novel graph representation of Web tables and lists based on a systematic categorization of the components in semi-structured data as well as their relations. We also develop pre-training and reasoning techniques on the graph model for the QA task. Extensive experiments on several real datasets collected from a commercial engine verify the effectiveness of our approach. Our method improves F1 score by 3.90 points over the state-of-the-art baselines.

Generating adversarial examples for natural language is hard, as natural language consists of discrete symbols, and examples are often of variable lengths. In this paper, we propose a geometry-inspired attack for generating natural language adversarial examples. Our attack generates adversarial examples by iteratively approximating the decision boundary of Deep Neural Networks (DNNs). Experiments on two datasets with two different models show that our attack fools natural language models with high success rates, while only replacing a few words. Human evaluation shows that adversarial examples generated by our attack are hard for humans to recognize. Further experiments show that adversarial training can improve model robustness against our attack.

Research on document-level Neural Machine Translation (NMT) models has attracted increasing attention in recent years. Although the proposed works have proved that the inter-sentence information is helpful for improving the performance of the NMT models, what information should be regarded as context remains ambiguous. To solve this problem, we proposed a novel cache-based document-level NMT model which conducts dynamic caching guided by theme-rheme information. The experiments on NIST evaluation sets demonstrate that our proposed model achieves substantial improvements over the state-of-the-art baseline NMT models. As far as we know, we are the first to introduce theme-rheme theory into the field of machine translation.

Passage retrieval is the task of identifying text snippets that are valid answers for a natural language posed question. One way to address this problem is to look at it as a metric learning problem, where we want to induce a metric between questions and passages that assign smaller distances to more relevant passages. In this work, we present a novel method for passage retrieval that learns a metric for questions and passages based on their internal semantic interactions. The method uses a similar approach to that of triplet networks, where the training samples are composed of one anchor (the question) and two positive and negative samples (passages). However,and in contrast with triplet networks, the proposed method uses a novel deep architecture that better exploits the particularities of text and takes into consideration complementary relatedness measures. Besides, the paper presents a sampling strategy that selects both easy and hard negative samples which improves the accuracy of the trained model. The method is particularly well suited for domain-specific passage retrieval where it is very important to take into account different sources of information. The proposed approach was evaluated in a biomedical passage retrieval task, the BioASQ challenge, outperforming standard triplet loss substantially by 10%,and state-of-the-art performance by 26%.