COLING 2020 论文列表

Offensive and abusive language is a pressing problem on social media platforms. In this work, we propose a method for transforming offensive comments, statements containing profanity or offensive language, into non-offensive ones. We design a Retrieve, Generate and Edit unsupervised style transfer pipeline to redact the offensive comments in a word-restricted manner while maintaining a high level of fluency and preserving the content of the original text. We extensively evaluate our method’s performance and compare it to previous style transfer models using both automatic metrics and human evaluations. Experimental results show that our method outperforms other models on human evaluations and is the only approach that consistently performs well on all automatic evaluation metrics.

Conventional neural generative models tend to generate safe and generic responses which have little connection with previous utterances semantically and would disengage users in a dialog system. To generate relevant responses, we propose a method that employs two types of constraints - topical constraint and semantic constraint. Under the hypothesis that a response and its context have higher relevance when they share the same topics, the topical constraint encourages the topics of a response to match its context by conditioning response decoding on topic words’ embeddings. The semantic constraint, which encourages a response to be semantically related to its context by regularizing the decoding objective function with semantic distance, is proposed. Optimal transport is applied to compute a weighted semantic distance between the representation of a response and the context. Generated responses are evaluated by automatic metrics, as well as human judgment, showing that the proposed method can generate more topic-relevant and content-rich responses than conventional models.

Online advertising is one of the most widespread ways to reach and increase a target audience for those selling products. Usually having a form of a banner, advertising engages users into visiting a corresponding webpage. Professional generation of banners requires creative and writing skills and a basic understanding of target products. The great variety of goods presented in the online market enforce professionals to spend more and more time creating new advertisements different from existing ones. In this paper, we propose a neural network-based approach for the automatic generation of online advertising using texts from given webpages as sources. The important part of the approach is training on open data available online, which allows avoiding costly procedures of manual labeling. Collected open data consist of multiple subdomains with high data heterogeneity. The subdomains belong to different topics and vary in used vocabularies, phrases, styles that lead to reduced quality in adverts generation. We try to solve the problem of identifying existed subdomains and proposing a new ensemble approach based on exploiting multiple instances of a seq2seq model. Our experimental study on a dataset in the Russian language shows that our approach can significantly improve the quality of adverts generation.

Neural machine translation with millions of parameters is vulnerable to unfamiliar inputs. We propose Token Drop to improve generalization and avoid overfitting for the NMT model. Similar to word dropout, whereas we replace dropped token with a special token instead of setting zero to words. We further introduce two self-supervised objectives: Replaced Token Detection and Dropped Token Prediction. Our method aims to force model generating target translation with less information, in this way the model can learn textual representation better. Experiments on Chinese-English and English-Romanian benchmark demonstrate the effectiveness of our approach and our model achieves significant improvements over a strong Transformer baseline.

Distantly Supervised Relation Extraction (DSRE) has proven to be effective to find relational facts from texts, but it still suffers from two main problems: the wrong labeling problem and the long-tail problem. Most of the existing approaches address these two problems through flat classification, which lacks hierarchical information of relations. To leverage the informative relation hierarchies, we formulate DSRE as a hierarchical classification task and propose a novel hierarchical classification framework, which extracts the relation in a top-down manner. Specifically, in our proposed framework, 1) we use a hierarchically-refined representation method to achieve hierarchy-specific representation; 2) a top-down classification strategy is introduced instead of training a set of local classifiers. The experiments on NYT dataset demonstrate that our approach significantly outperforms other state-of-the-art approaches, especially for the long-tail problem.

This paper addresses the question as to what degree a BERT-based multilingual Spoken Language Understanding (SLU) model can transfer knowledge across languages. Through experiments we will show that, although it works substantially well even on distant language groups, there is still a gap to the ideal multilingual performance. In addition, we propose a novel BERT-based adversarial model architecture to learn language-shared and language-specific representations for multilingual SLU. Our experimental results prove that the proposed model is capable of narrowing the gap to the ideal multilingual performance.

We reduce the model size of pre-trained word embeddings by a factor of 200 while preserving its quality. Previous studies in this direction created a smaller word embedding model by reconstructing pre-trained word representations from those of subwords, which allows to store only a smaller number of subword embeddings in the memory. However, previous studies that train the reconstruction models using only target words cannot reduce the model size extremely while preserving its quality. Inspired by the observation of words with similar meanings having similar embeddings, our reconstruction training learns the global relationships among words, which can be employed in various models for word embedding reconstruction. Experimental results on word similarity benchmarks show that the proposed method improves the performance of the all subword-based reconstruction models.

Dubbing has two shades; synchronisation constraints are applied only when the actor’s mouth is visible on screen, while the translation is unconstrained for off-screen dubbing. Consequently, different synchronisation requirements, and therefore translation strategies, are applied depending on the type of dubbing. In this work, we manually annotate an existing dubbing corpus (Heroes) for this dichotomy. We show that, even though we did not observe distinctive features between on- and off-screen dubbing at the textual level, on-screen dubbing is more difficult for MT (-4 BLEU points). Moreover, synchronisation constraints dramatically decrease translation quality for off-screen dubbing. We conclude that, distinguishing between on-screen and off-screen dubbing is necessary for determining successful strategies for dubbing-customised Machine Translation.

In automatic post-editing (APE) it makes sense to condition post-editing (pe) decisions on both the source (src) and the machine translated text (mt) as input. This has led to multi-encoder based neural APE approaches. A research challenge now is the search for architectures that best support the capture, preparation and provision of src and mt information and its integration with pe decisions. In this paper we present an efficient multi-encoder based APE model, called transference. Unlike previous approaches, it (i) uses a transformer encoder block for src, (ii) followed by a decoder block, but without masking for self-attention on mt, which effectively acts as second encoder combining src –> mt, and (iii) feeds this representation into a final decoder block generating pe. Our model outperforms the best performing systems by 1 BLEU point on the WMT 2016, 2017, and 2018 English–German APE shared tasks (PBSMT and NMT). Furthermore, the results of our model on the WMT 2019 APE task using NMT data shows a comparable performance to the state-of-the-art system. The inference time of our model is similar to the vanilla transformer-based NMT system although our model deals with two separate encoders. We further investigate the importance of our newly introduced second encoder and find that a too small amount of layers does hurt the performance, while reducing the number of layers of the decoder does not matter much.

Low-resource Machine Translation recently gained a lot of popularity, and for certain languages, it has made great strides. However, it is still difficult to track progress in other languages for which there is no publicly available evaluation data. In this paper, we introduce benchmark datasets for Arabic and its dialects. We describe our design process and motivations and analyze the datasets to understand their resulting properties. Numerous successful attempts use large monolingual corpora to augment low-resource pairs. We try to approach augmentation differently and investigate whether it is possible to improve MT models without any external sources of data. We accomplish this by bootstrapping existing parallel sentences and complement this with multilingual training to achieve strong baselines.

This paper surveys the first, three-year phase of a project at the National Research Council of Canada that is developing software to assist Indigenous communities in Canada in preserving their languages and extending their use. The project aimed to work within the empowerment paradigm, where collaboration with communities and fulfillment of their goals is central. Since many of the technologies we developed were in response to community needs, the project ended up as a collection of diverse subprojects, including the creation of a sophisticated framework for building verb conjugators for highly inflectional polysynthetic languages (such as Kanyen’kéha, in the Iroquoian language family), release of what is probably the largest available corpus of sentences in a polysynthetic language (Inuktut) aligned with English sentences and experiments with machine translation (MT) systems trained on this corpus, free online services based on automatic speech recognition (ASR) for easing the transcription bottleneck for recordings of speech in Indigenous languages (and other languages), software for implementing text prediction and read-along audiobooks for Indigenous languages, and several other subprojects.

Contextual embeddings derived from transformer-based neural language models have shown state-of-the-art performance for various tasks such as question answering, sentiment analysis, and textual similarity in recent years. Extensive work shows how accurately such models can represent abstract, semantic information present in text. In this expository work, we explore a tangent direction and analyze such models’ performance on tasks that require a more granular level of representation. We focus on the problem of textual similarity from two perspectives: matching documents on a granular level (requiring embeddings to capture fine-grained attributes in the text), and an abstract level (requiring embeddings to capture overall textual semantics). We empirically demonstrate, across two datasets from different domains, that despite high performance in abstract document matching as expected, contextual embeddings are consistently (and at times, vastly) outperformed by simple baselines like TF-IDF for more granular tasks. We then propose a simple but effective method to incorporate TF-IDF into models that use contextual embeddings, achieving relative improvements of up to 36% on granular tasks.

News articles, image captions, product reviews and many other texts mention people and organizations whose name recognition could vary for different audiences. In such cases, background information about the named entities could be provided in the form of an appositive noun phrase, either written by a human or generated automatically. We expand on the previous work in appositive generation with a new, more realistic, end-to-end definition of the task, instantiated by a dataset that spans four languages (English, Spanish, German and Polish), two entity types (person and organization) and two domains (Wikipedia and News). We carry out an extensive analysis of the data and the task, pointing to the various modeling challenges it poses. The results we obtain with standard language generation methods show that the task is indeed non-trivial, and leaves plenty of room for improvement.

Data scarcity is always a constraint on analyzing speech transcriptions for automatic Alzheimer’s disease (AD) detection, especially when the subjects are non-English speakers. To deal with this issue, this paper first proposes a contrastive learning method to obtain effective representations for text classification based on monolingual embeddings of BERT. Furthermore, a cross-lingual data augmentation method is designed by building autoencoders to learn the text representations shared by both languages. Experiments on a Mandarin AD corpus show that the contrastive learning method can achieve better detection accuracy than conventional CNN-based and BERTbased methods. Our cross-lingual data augmentation method also outperforms other compared methods when using another English AD corpus for augmentation. Finally, a best detection accuracy of 81.6% is obtained by our proposed methods on the Mandarin AD corpus.

The computational treatment of temporal relations is based on the work of Allen, who establishes 13 different types, and Freksa, who designs a cognitive procedure to manage them. Freksa’s notation is not widely used because, although it has cognitive and expressive advantages, it is too complex from the computational perspective. This paper proposes a system for the annotation and management of temporal relations that combines the richness and expressiveness of Freksa’s approach with the simplicity of Allen’s notation. Our method is summarized in the application of bounding relations, thanks to which it is possible to obtain the temporary representation of complete neighborhoods capable of representing vague temporal relations such as those that can be frequently found in a text. Such advantages are obtained without the need to greatly increase the complexity of the labeling process since the markup language is almost the same as TimeML, to which only a second temporary “relType”’ type label relationship is added. Our experiments show that the temporal relationships that present vagueness are in fact much more common than those in which a single relationship can be established precisely. For these reasons, our new labeling system achieves a more agreeable representation of temporal relations.

In the last few years, there has been a surge of interest in learning representations of entities and relations in knowledge graph (KG). However, the recent availability of temporal knowledge graphs (TKGs) that contain time information for each fact created the need for reasoning over time in such TKGs. In this regard, we present a new approach of TKG embedding, TeRo, which defines the temporal evolution of entity embedding as a rotation from the initial time to the current time in the complex vector space. Specially, for facts involving time intervals, each relation is represented as a pair of dual complex embeddings to handle the beginning and the end of the relation, respectively. We show our proposed model overcomes the limitations of the existing KG embedding models and TKG embedding models and has the ability of learning and inferring various relation patterns over time. Experimental results on three different TKGs show that TeRo significantly outperforms existing state-of-the-art models for link prediction. In addition, we analyze the effect of time granularity on link prediction over TKGs, which as far as we know has not been investigated in previous literature.

State-of-the-art approaches for text classification leverage a transformer architecture with a linear layer on top that outputs a class distribution for a given prediction problem. While effective, this approach suffers from conceptual limitations that affect its utility in few-shot or zero-shot transfer learning scenarios. First, the number of classes to predict needs to be pre-defined. In a transfer learning setting, in which new classes are added to an already trained classifier, all information contained in a linear layer is therefore discarded, and a new layer is trained from scratch. Second, this approach only learns the semantics of classes implicitly from training examples, as opposed to leveraging the explicit semantic information provided by the natural language names of the classes. For instance, a classifier trained to predict the topics of news articles might have classes like “business” or “sports” that themselves carry semantic information. Extending a classifier to predict a new class named “politics” with only a handful of training examples would benefit from both leveraging the semantic information in the name of a new class and using the information contained in the already trained linear layer. This paper presents a novel formulation of text classification that addresses these limitations. It imbues the notion of the task at hand into the transformer model itself by factorizing arbitrary classification problems into a generic binary classification problem. We present experiments in few-shot and zero-shot transfer learning that show that our approach significantly outperforms previous approaches on small training data and can even learn to predict new classes with no training examples at all. The implementation of our model is publicly available at: https://github.com/flairNLP/flair.

Existing metonymy resolution approaches rely on features extracted from external resources like dictionaries and hand-crafted lexical resources. In this paper, we propose an end-to-end word-level classification approach based only on BERT, without dependencies on taggers, parsers, curated dictionaries of place names, or other external resources. We show that our approach achieves the state-of-the-art on 5 datasets, surpassing conventional BERT models and benchmarks by a large margin. We also show that our approach generalises well to unseen data.

This paper presents performance measures for grammatical error correction which take into account the difficulty of error correction. To the best of our knowledge, no conventional measure has such functionality despite the fact that some errors are easy to correct and others are not. The main purpose of this work is to provide a way of determining the difficulty of error correction and to motivate researchers in the domain to attack such difficult errors. The performance measures are based on the simple idea that the more systems successfully correct an error, the easier it is considered to be. This paper presents a set of algorithms to implement this idea. It evaluates the performance measures quantitatively and qualitatively on a wide variety of corpora and systems, revealing that they agree with our intuition of correction difficulty. A scorer and difficulty weight data based on the algorithms have been made available on the web.