NAACL 2022 Papers

Target-guided response generation enables dialogue systems to smoothly transition a conversation from a dialogue context toward a target sentence. Such control is useful for designing dialogue systems that direct a conversation toward specific goals, such as creating non-obtrusive recommendations or introducing new topics in the conversation. In this paper, we introduce a new technique for target-guided response generation, which first finds a bridging path of commonsense knowledge concepts between the source and the target, and then uses the identified bridging path to generate transition responses. Additionally, we propose techniques to re-purpose existing dialogue datasets for target-guided generation. Experiments reveal that the proposed techniques outperform various baselines on this task. Finally, we observe that the existing automated metrics for this task correlate poorly with human judgement ratings. We propose a novel evaluation metric that we demonstrate is more reliable for target-guided response evaluation. Our work generally enables dialogue system designers to exercise more control over the conversations that their systems produce.

Masked language models (MLMs) such as BERT have revolutionized the field of Natural Language Understanding in the past few years. However, existing pre-trained MLMs often output an anisotropic distribution of token representations that occupies a narrow subset of the entire representation space. Such token representations are not ideal, especially for tasks that demand discriminative semantic meanings of distinct tokens. In this work, we propose TaCL (Token-aware Contrastive Learning), a novel continual pre-training approach that encourages BERT to learn an isotropic and discriminative distribution of token representations. TaCL is fully unsupervised and requires no additional data. We extensively test our approach on a wide range of English and Chinese benchmarks. The results show that TaCL brings consistent and notable improvements over the original BERT model. Furthermore, we conduct detailed analysis to reveal the merits and inner-workings of our approach.

Recently, several studies on propaganda detection have involved document and fragment-level analyses of news articles. However, there are significant data and modeling challenges dealing with fine-grained detection of propaganda on social media. In this work, we present TWEETSPIN, a dataset containing tweets that are weakly annotated with different fine-grained propaganda techniques, and propose a neural approach to detect and categorize propaganda tweets across those fine-grained categories. These categories include specific rhetorical and psychological techniques, ranging from leveraging emotions to using logical fallacies. Our model relies on multi-view representations of the input tweet data to (a) extract different aspects of the input text including the context, entities, their relationships, and external knowledge; (b) model their mutual interplay; and (c) effectively speed up the learning process by requiring fewer training examples. Our method allows for representation enrichment leading to better detection and categorization of propaganda on social media. We verify the effectiveness of our proposed method on TWEETSPIN and further probe how the implicit relations between the views impact the performance. Our experiments show that our model is able to outperform several benchmark methods and transfer the knowledge to relatively low-resource news domains.

We propose a new task for assessing machines’ skills of understanding fictional characters in narrative stories. The task, TVShowGuess, builds on the scripts of TV series and takes the form of guessing the anonymous main characters based on the backgrounds of the scenes and the dialogues. Our human study supports that this form of task covers comprehension of multiple types of character persona, including understanding characters’ personalities, facts and memories of personal experience, which are well aligned with the psychological and literary theories about the theory of mind (ToM) of human beings on understanding fictional characters during reading. We further propose new model architectures to support the contextualized encoding of long scene texts. Experiments show that our proposed approaches significantly outperform baselines, yet still largely lag behind the (nearly perfect) human performance. Our work serves as a first step toward the goal of narrative character comprehension.

Many scientific papers such as those in arXiv and PubMed data collections have abstracts with varying lengths of 50-1000 words and average length of approximately 200 words, where longer abstracts typically convey more information about the source paper. Up to recently, scientific summarization research has typically focused on generating short, abstract-like summaries following the existing datasets used for scientific summarization. In domains where the source text is relatively long-form, such as in scientific documents, such summary is not able to go beyond the general and coarse overview and provide salient information from the source document. The recent interest to tackle this problem motivated curation of scientific datasets, arXiv-Long and PubMed-Long, containing human-written summaries of 400-600 words, hence, providing a venue for research in generating long/extended summaries. Extended summaries facilitate a faster read while providing details beyond coarse information. In this paper, we propose TSTR, an extractive summarizer that utilizes the introductory information of documents as pointers to their salient information. The evaluations on two existing large-scale extended summarization datasets indicate statistically significant improvement in terms of Rouge and average Rouge (F1) scores (except in one case) as compared to strong baselines and state-of-the-art. Comprehensive human evaluations favor our generated extended summaries in terms of cohesion and completeness.

Grounded text generation systems often generate text that contains factual inconsistencies, hindering their real-world applicability. Automatic factual consistency evaluation may help alleviate this limitation by accelerating evaluation cycles, filtering inconsistent outputs and augmenting training data. While attracting increasing attention, such evaluation metrics are usually developed and evaluated in silo for a single task or dataset, slowing their adoption. Moreover, previous meta-evaluation protocols focused on system-level correlations with human annotations, which leave the example-level accuracy of such metrics unclear. In this work, we introduce TRUE: a comprehensive survey and assessment of factual consistency metrics on a standardized collection of existing texts from diverse tasks, manually annotated for factual consistency. Our standardization enables an example-level meta-evaluation protocol that is more actionable and interpretable than previously reported correlations, yielding clearer quality measures. Across diverse state-of-the-art metrics and 11 datasets we find that large-scale NLI and question generation-and-answering-based approaches achieve strong and complementary results. We recommend those methods as a starting point for model and metric developers, and hope TRUE will foster progress towards even better evaluation methods.

Recently, the structural reading comprehension (SRC) task on web pages has attracted increasing research interests. Although previous SRC work has leveraged extra information such as HTML tags or XPaths, the informative topology of web pages is not effectively exploited. In this work, we propose a Topological Information Enhanced model (TIE), which transforms the token-level task into a tag-level task by introducing a two-stage process (i.e. node locating and answer refining). Based on that, TIE integrates Graph Attention Network (GAT) and Pre-trained Language Model (PLM) to leverage the topological information of both logical structures and spatial structures. Experimental results demonstrate that our model outperforms strong baselines and achieves state-of-the-art performances on the web-based SRC benchmark WebSRC at the time of writing. The code of TIE will be publicly available at https://github.com/X-LANCE/TIE.

Taxonomy expansion is a crucial task. Most of Automatic expansion of taxonomy are of two types, attach and merge. In a taxonomy like WordNet, both merge and attach are integral parts of the expansion operations but majority of study consider them separately. This paper proposes a novel mult-task learning-based deep learning method known as Taxonomy Expansion with Attach and Merge (TEAM) that performs both the merge and attach operations. To the best of our knowledge this is the first study which integrates both merge and attach operations in a single model. The proposed models have been evaluated on three separate WordNet taxonomies, viz., Assamese, Bangla, and Hindi. From the various experimental setups, it is shown that TEAM outperforms its state-of-the-art counterparts for attach operation, and also provides highly encouraging performance for the merge operation.

Although pre-trained language models (PLMs) have achieved great success and become a milestone in NLP, abstractive conversational summarization remains a challenging but less studied task. The difficulty lies in two aspects. One is the lack of large-scale conversational summary data. Another is that applying the existing pre-trained models to this task is tricky because of the structural dependence within the conversation and its informal expression, etc. In this work, we first build a large-scale (11M) pretraining dataset called RCSum, based on the multi-person discussions in the Reddit community. We then present TANet, a thread-aware Transformer-based network. Unlike the existing pre-trained models that treat a conversation as a sequence of sentences, we argue that the inherent contextual dependency among the utterances plays an essential role in understanding the entire conversation and thus propose two new techniques to incorporate the structural information into our model. The first is thread-aware attention which is computed by taking into account the contextual dependency within utterances. Second, we apply thread prediction loss to predict the relations between utterances. We evaluate our model on four datasets of real conversations, covering types of meeting transcripts, customer-service records, and forum threads. Experimental results demonstrate that TANet achieves a new state-of-the-art in terms of both automatic evaluation and human judgment.

The knowledge graph (KG) stores a large amount of structural knowledge, while it is not easy for direct human understanding. Knowledge graph-to-text (KG-to-text) generation aims to generate easy-to-understand sentences from the KG, and at the same time, maintains semantic consistency between generated sentences and the KG. Existing KG-to-text generation methods phrase this task as a sequence-to-sequence generation task with linearized KG as input and consider the consistency issue of the generated texts and KG through a simple selection between decoded sentence word and KG node word at each time step. However, the linearized KG order is obtained through a heuristic search without data-driven optimization. In this paper, we optimize the knowledge description order prediction under the order supervision extracted from the caption and further enhance the consistency of the generated sentences and KG through syntactic and semantic regularization. We incorporate the Part-of-Speech (POS) syntactic tags to constrain the positions to copy words from the KG and employ a semantic context scoring function to evaluate the semantic fitness for each word in its local context when decoding each word in the generated sentence. Extensive experiments are conducted on two datasets, WebNLG and DART, and achieve state-of-the-art performances. Our code is now public available.

In this paper we focus on patterns of colexification (co-expressions of form-meaning mapping in the lexicon) as an aspect of lexical-semantic organization, and use them to build large scale synset graphs across BabelNet’s typologically diverse set of 499 world languages. We introduce and compare several approaches: monolingual and cross-lingual colexification graphs, popular distributional models, and fusion approaches. The models are evaluated against human judgments on a semantic similarity task for nine languages. Our strong empirical findings also point to the importance of universality of our graph synset embedding representations with no need for any language-specific adaptation when evaluated on the lexical similarity task. The insights of our exploratory investigation of large-scale colexification graphs could inspire significant advances in NLP across languages, especially for tasks involving languages which lack dedicated lexical resources, and can benefit from language transfer from large shared cross-lingual semantic spaces.

The common practice for training commonsense models has gone from–human–to–corpus–to–machine: humans author commonsense knowledge graphs in order to train commonsense models. In this work, we investigate an alternative, from–machine–to–corpus–to–machine: general language models author these commonsense knowledge graphs to train commonsense models. Our study leads to a new framework, Symbolic Knowledge Distillation. As with prior art in Knowledge Distillation (Hinton et al. 2015), our approach uses larger models to teach smaller models. A key difference is that we distill knowledge symbolically–as text–in addition to the neural model. We distill only one aspect–the commonsense of a general language model teacher, allowing the student to be a different type, a commonsense model. Altogether, we show that careful prompt engineering and a separately trained critic model allow us to selectively distill high-quality causal commonsense from GPT-3, a general language model. Empirical results demonstrate that, for the first time, a human-authored commonsense knowledge graph is surpassed by our automatically distilled variant in all three criteria: quantity, quality, and diversity. In addition, it results in a neural commonsense model that surpasses the teacher model’s commonsense capabilities despite its 100x smaller size. We apply this to the ATOMIC resource, and will share our new symbolic knowledge graph and commonsense models.

The rapid development of social networks, electronic commerce, mobile Internet, and other technologies, has influenced the growth of Web data. Social media and Internet forums are valuable sources of citizens’ opinions, which can be analyzed for community development and user behavior analysis. Unfortunately, the scarcity of resources (i.e., datasets or language models) become a barrier to the development of natural language processing applications in low-resource languages. Thanks to the recent growth of online forums and news platforms of Swahili, we introduce two datasets of Swahili in this paper: a pre-training dataset of approximately 105MB with 16M words and annotated dataset of 13K instances for the emotion classification task. The emotion classification dataset is manually annotated by two native Swahili speakers. We pre-trained a new monolingual language model for Swahili, namely SwahBERT, using our collected pre-training data, and tested it with four downstream tasks including emotion classification. We found that SwahBERT outperforms multilingual BERT, a well-known existing language model, in almost all downstream tasks.

Canonical automatic summary evaluation metrics, such as ROUGE, focus on lexical similarity which cannot well capture semantics nor linguistic quality and require a reference summary which is costly to obtain. Recently, there have been a growing number of efforts to alleviate either or both of the two drawbacks. In this paper, we present a proof-of-concept study to a weakly supervised summary evaluation approach without the presence of reference summaries. Massive data in existing summarization datasets are transformed for training by pairing documents with corrupted reference summaries. In cross-domain tests, our strategy outperforms baselines with promising improvements, and show a great advantage in gauging linguistic qualities over all metrics.

Current Knowledge-Grounded Dialogue Generation (KDG) models specialize in producing rational and factual responses. However, to establish long-term relationships with users, the KDG model needs the capability to generate responses in a desired style or attribute. Thus, we study a new problem: Stylized Knowledge-Grounded Dialogue Generation (SKDG). It presents two challenges: (1) How to train a SKDG model where no <context, knowledge, stylized response> triples are available. (2) How to cohere with context and preserve the knowledge when generating a stylized response. In this paper, we propose a novel disentangled template rewriting (DTR) method which generates responses via combing disentangled style templates (from monolingual stylized corpus) and content templates (from KDG corpus). The entire framework is end-to-end differentiable and learned without supervision. Extensive experiments on two benchmarks indicate that DTR achieves a significant improvement on all evaluation metrics compared with previous state-of-the-art stylized dialogue generation methods. Besides, DTR achieves comparable performance with the state-of-the-art KDG methods in standard KDG evaluation setting.

Knowledge graphs (KGs) often represent knowledge bases that are incomplete. Machine learning models can alleviate this by helping automate graph completion. Recently, there has been growing interest in completing knowledge bases that are dynamic, where previously unseen entities may be added to the KG with many missing links. In this paper, we present StATIK–Structure And Text for Inductive Knowledge Completion. StATIK uses Language Models to extract the semantic information from text descriptions, while using Message Passing Neural Networks to capture the structural information. StATIK achieves state of the art results on three challenging inductive baselines. We further analyze our hybrid model through detailed ablation studies.

Softmax is the de facto standard for normalizing logits in modern neural networks for language processing. However, by producing a dense probability distribution each token in the vocabulary has a nonzero chance of being selected at each generation step, leading to a variety of reported problems in text generation. 𝛼-entmax of Peters et al. (2019) solves this problem, but is unfortunately slower than softmax. In this paper, we propose an alternative to 𝛼-entmax, which keeps its virtuous characteristics, but is as fast as optimized softmax and achieves on par or better performance in machine translation task.

Pretrained masked language models (PLMs) were shown to be inheriting a considerable amount of relational knowledge from the source corpora. In this paper, we present an in-depth and comprehensive study concerning specializing PLMs into relational models from the perspective of network pruning. We show that it is possible to find subnetworks capable of representing grounded commonsense relations at non-trivial sparsity while being more generalizable than original PLMs in scenarios requiring knowledge of single or multiple commonsense relations.

Distilling state-of-the-art transformer models into lightweight student models is an effective way to reduce computation cost at inference time. The student models are typically compact transformers with fewer parameters, while expensive operations such as self-attention persist. Therefore, the improved inference speed may still be unsatisfactory for real-time or high-volume use cases. In this paper, we aim to further push the limit of inference speed by distilling teacher models into bigger, sparser student models – bigger in that they scale up to billions of parameters; sparser in that most of the model parameters are n-gram embeddings. Our experiments on six single-sentence text classification tasks show that these student models retain 97% of the RoBERTa-Large teacher performance on average, and meanwhile achieve up to 600x speed-up on both GPUs and CPUs at inference time. Further investigation reveals that our pipeline is also helpful for sentence-pair classification tasks, and in domain generalization settings.