EMNLP 2021 Papers

Knowledge Base Question Answering (KBQA) is to answer natural language questions posed over knowledge bases (KBs). This paper targets at empowering the IR-based KBQA models with the ability of numerical reasoning for answering ordinal constrained questions. A major challenge is the lack of explicit annotations about numerical properties. To address this challenge, we propose a pretraining numerical reasoning model consisting of NumGNN and NumTransformer, guided by explicit self-supervision signals. The two modules are pretrained to encode the magnitude and ordinal properties of numbers respectively and can serve as model-agnostic plugins for any IR-based KBQA model to enhance its numerical reasoning ability. Extensive experiments on two KBQA benchmarks verify the effectiveness of our method to enhance the numerical reasoning ability for IR-based KBQA models.

Large pre-trained language models have repeatedly shown their ability to produce fluent text. Yet even when starting from a prompt, generation can continue in many plausible directions. Current decoding methods with the goal of controlling generation, e.g., to ensure specific words are included, either require additional models or fine-tuning, or work poorly when the task at hand is semantically unconstrained, e.g., story generation. In this work, we present a plug-and-play decoding method for controlled language generation that is so simple and intuitive, it can be described in a single sentence: given a topic or keyword, we add a shift to the probability distribution over our vocabulary towards semantically similar words. We show how annealing this distribution can be used to impose hard constraints on language generation, something no other plug-and-play method is currently able to do with SOTA language generators. Despite the simplicity of this approach, we see it works incredibly well in practice: decoding from GPT-2 leads to diverse and fluent sentences while guaranteeing the appearance of given guide words. We perform two user studies, revealing that (1) our method outperforms competing methods in human evaluations; and (2) forcing the guide words to appear in the generated text has no impact on the fluency of the generated text.

In joint entity and relation extraction, existing work either sequentially encode task-specific features, leading to an imbalance in inter-task feature interaction where features extracted later have no direct contact with those that come first. Or they encode entity features and relation features in a parallel manner, meaning that feature representation learning for each task is largely independent of each other except for input sharing. We propose a partition filter network to model two-way interaction between tasks properly, where feature encoding is decomposed into two steps: partition and filter. In our encoder, we leverage two gates: entity and relation gate, to segment neurons into two task partitions and one shared partition. The shared partition represents inter-task information valuable to both tasks and is evenly shared across two tasks to ensure proper two-way interaction. The task partitions represent intra-task information and are formed through concerted efforts of both gates, making sure that encoding of task-specific features is dependent upon each other. Experiment results on six public datasets show that our model performs significantly better than previous approaches. In addition, contrary to what previous work has claimed, our auxiliary experiments suggest that relation prediction is contributory to named entity prediction in a non-negligible way. The source code can be found at https://github.com/Coopercoppers/PFN.

Table filling based relational triple extraction methods are attracting growing research interests due to their promising performance and their abilities on extracting triples from complex sentences. However, this kind of methods are far from their full potential because most of them only focus on using local features but ignore the global associations of relations and of token pairs, which increases the possibility of overlooking some important information during triple extraction. To overcome this deficiency, we propose a global feature-oriented triple extraction model that makes full use of the mentioned two kinds of global associations. Specifically, we first generate a table feature for each relation. Then two kinds of global associations are mined from the generated table features. Next, the mined global associations are integrated into the table feature of each relation. This “generate-mine-integrate” process is performed multiple times so that the table feature of each relation is refined step by step. Finally, each relation’s table is filled based on its refined table feature, and all triples linked to this relation are extracted based on its filled table. We evaluate the proposed model on three benchmark datasets. Experimental results show our model is effective and it achieves state-of-the-art results on all of these datasets. The source code of our work is available at: https://github.com/neukg/GRTE.

This paper proposes a new representation for CCG derivations. CCG derivations are represented as trees whose nodes are labeled with categories strictly restricted by CCG rule schemata. This characteristic is not suitable for span-based parsing models because they predict node labels independently. In other words, span-based models may generate invalid CCG derivations that violate the rule schemata. Our proposed representation decomposes CCG derivations into several independent pieces and prevents the span-based parsing models from violating the schemata. Our experimental result shows that an off-the-shelf span-based parser with our representation is comparable with previous CCG parsers.

Entity grids and entity graphs are two frameworks for modeling local coherence. These frameworks represent entity relations between sentences and then extract features from such representations to encode coherence. The benefits of convolutional neural models for extracting informative features from entity grids have been recently studied. In this work, we study the benefits of Relational Graph Convolutional Networks (RGCN) to encode entity graphs for measuring local coherence. We evaluate our neural graph-based model for two benchmark coherence evaluation tasks: sentence ordering (SO) and summary coherence rating (SCR). The results show that our neural graph-based model consistently outperforms the neural grid-based model for both tasks. Our model performs competitively with a strong baseline coherence model, while our model uses 50% fewer parameters. Our work defines a new, efficient, and effective baseline for local coherence modeling.

Multi-hop relation detection in Knowledge Base Question Answering (KBQA) aims at retrieving the relation path starting from the topic entity to the answer node based on a given question, where the relation path may comprise multiple relations. Most of the existing methods treat it as a single-label learning problem while ignoring the fact that for some complex questions, there exist multiple correct relation paths in knowledge bases. Therefore, in this paper, multi-hop relation detection is considered as a multi-label learning problem. However, performing multi-label multi-hop relation detection is challenging since the numbers of both the labels and the hops are unknown. To tackle this challenge, multi-label multi-hop relation detection is formulated as a sequence generation task. A relation-aware sequence relation generation model is proposed to solve the problem in an end-to-end manner. Experimental results show the effectiveness of the proposed method for relation detection and KBQA.

Research on open-domain dialogue systems that allow free topics is challenging in the field of natural language processing (NLP). The performance of the dialogue system has been improved recently by the method utilizing dialogue-related knowledge; however, non-English dialogue systems suffer from reproducing the performance of English dialogue systems because securing knowledge in the same language with the dialogue system is relatively difficult. Through experiments with a Korean dialogue system, this paper proves that the performance of a non-English dialogue system can be improved by utilizing English knowledge, highlighting the system uses cross-lingual knowledge. For the experiments, we 1) constructed a Korean version of the Wizard of Wikipedia dataset, 2) built Korean-English T5 (KE-T5), a language model pre-trained with Korean and English corpus, and 3) developed a knowledge-grounded Korean dialogue model based on KE-T5. We observed the performance improvement in the open-domain Korean dialogue model even only English knowledge was given. The experimental results showed that the knowledge inherent in cross-lingual language models can be helpful for generating responses in open dialogue systems.

In cross-lingual language models, representations for many different languages live in the same space. Here, we investigate the linguistic and non-linguistic factors affecting sentence-level alignment in cross-lingual pretrained language models for 101 languages and 5,050 language pairs. Using BERT-based LaBSE and BiLSTM-based LASER as our models, and the Bible as our corpus, we compute a task-based measure of cross-lingual alignment in the form of bitext retrieval performance, as well as four intrinsic measures of vector space alignment and isomorphism. We then examine a range of linguistic, quasi-linguistic, and training-related features as potential predictors of these alignment metrics. The results of our analyses show that word order agreement and agreement in morphological complexity are two of the strongest linguistic predictors of cross-linguality. We also note in-family training data as a stronger predictor than language-specific training data across the board. We verify some of our linguistic findings by looking at the effect of morphological segmentation on English-Inuktitut alignment, in addition to examining the effect of word order agreement on isomorphism for 66 zero-shot language pairs from a different corpus. We make the data and code for our experiments publicly available.

Recent advances in neural machine translation (NMT) have pushed the quality of machine translation systems to the point where they are becoming widely adopted to build competitive systems. However, there is still a large number of languages that are yet to reap the benefits of NMT. In this paper, we provide the first large-scale case study of the practical application of MT in the Turkic language family in order to realize the gains of NMT for Turkic languages under high-resource to extremely low-resource scenarios. In addition to presenting an extensive analysis that identifies the bottlenecks towards building competitive systems to ameliorate data scarcity, our study has several key contributions, including, i) a large parallel corpus covering 22 Turkic languages consisting of common public datasets in combination with new datasets of approximately 1.4 million parallel sentences, ii) bilingual baselines for 26 language pairs, iii) novel high-quality test sets in three different translation domains and iv) human evaluation scores. All models, scripts, and data will be released to the public.

Recent development in NLP shows a strong trend towards refining pre-trained models with a domain-specific dataset. This is especially the case for response generation where emotion plays an important role. However, existing empathetic datasets remain small, delaying research efforts in this area, for example, the development of emotion-aware chatbots. One main technical challenge has been the cost of manually annotating dialogues with the right emotion labels. In this paper, we describe a large-scale silver dataset consisting of 1M dialogues annotated with 32 fine-grained emotions, eight empathetic response intents, and the Neutral category. To achieve this goal, we have developed a novel data curation pipeline starting with a small seed of manually annotated data and eventually scaling it to a satisfactory size. We compare its quality against a state-of-the-art gold dataset using both offline experiments and visual validation methods. The resultant procedure can be used to create similar datasets in the same domain as well as in other domains.

Discourse segmentation and sentence-level discourse parsing play important roles for various NLP tasks to consider textual coherence. Despite recent achievements in both tasks, there is still room for improvement due to the scarcity of labeled data. To solve the problem, we propose a language model-based generative classifier (LMGC) for using more information from labels by treating the labels as an input while enhancing label representations by embedding descriptions for each label. Moreover, since this enables LMGC to make ready the representations for labels, unseen in the pre-training step, we can effectively use a pre-trained language model in LMGC. Experimental results on the RST-DT dataset show that our LMGC achieved the state-of-the-art F1 score of 96.72 in discourse segmentation. It further achieved the state-of-the-art relation F1 scores of 84.69 with gold EDU boundaries and 81.18 with automatically segmented boundaries, respectively, in sentence-level discourse parsing.

With the early success of query-answer assistants such as Alexa and Siri, research attempts to expand system capabilities of handling service automation are now abundant. However, preliminary systems have quickly found the inadequacy in relying on simple classification techniques to effectively accomplish the automation task. The main challenge is that the dialogue often involves complexity in user’s intents (or purposes) which are multiproned, subject to spontaneous change, and difficult to track. Furthermore, public datasets have not considered these complications and the general semantic annotations are lacking which may result in zero-shot problem. Motivated by the above, we propose a Label-Aware BERT Attention Network (LABAN) for zero-shot multi-intent detection. We first encode input utterances with BERT and construct a label embedded space by considering embedded semantics in intent labels. An input utterance is then classified based on its projection weights on each intent embedding in this embedded space. We show that it successfully extends to few/zero-shot setting where part of intent labels are unseen in training data, by also taking account of semantics in these unseen intent labels. Experimental results show that our approach is capable of detecting many unseen intent labels correctly. It also achieves the state-of-the-art performance on five multi-intent datasets in normal cases.

Frame semantic parsing is a semantic analysis task based on FrameNet which has received great attention recently. The task usually involves three subtasks sequentially: (1) target identification, (2) frame classification and (3) semantic role labeling. The three subtasks are closely related while previous studies model them individually, which ignores their intern connections and meanwhile induces error propagation problem. In this work, we propose an end-to-end neural model to tackle the task jointly. Concretely, we exploit a graph-based method, regarding frame semantic parsing as a graph construction problem. All predicates and roles are treated as graph nodes, and their relations are taken as graph edges. Experiment results on two benchmark datasets of frame semantic parsing show that our method is highly competitive, resulting in better performance than pipeline models.

We propose a generative framework for simultaneous machine translation. Conventional approaches use a fixed number of source words to translate or learn dynamic policies for the number of source words by reinforcement learning. Here we formulate simultaneous translation as a structural sequence-to-sequence learning problem. A latent variable is introduced to model read or translate actions at every time step, which is then integrated out to consider all the possible translation policies. A re-parameterised Poisson prior is used to regularise the policies which allows the model to explicitly balance translation quality and latency. The experiments demonstrate the effectiveness and robustness of the generative framework, which achieves the best BLEU scores given different average translation latencies on benchmark datasets.

Although abstractive summarization models have achieved impressive results on document summarization tasks, their performance on dialogue modeling is much less satisfactory due to the crude and straight methods for dialogue encoding. To address this question, we propose a novel end-to-end Transformer-based model FinDS for abstractive dialogue summarization that leverages Finer-grain universal Dialogue semantic Structures to model dialogue and generates better summaries. Experiments on the SAMsum dataset show that FinDS outperforms various dialogue summarization approaches and achieves new state-of-the-art (SOTA) ROUGE results. Finally, we apply FinDS to a more complex scenario, showing the robustness of our model. We also release our source code.

Domain adaption for word segmentation and POS tagging is a challenging problem for Chinese lexical processing. Self-training is one promising solution for it, which struggles to construct a set of high-quality pseudo training instances for the target domain. Previous work usually assumes a universal source-to-target adaption to collect such pseudo corpus, ignoring the different gaps from the target sentences to the source domain. In this work, we start from joint word segmentation and POS tagging, presenting a fine-grained domain adaption method to model the gaps accurately. We measure the gaps by one simple and intuitive metric, and adopt it to develop a pseudo target domain corpus based on fine-grained subdomains incrementally. A novel domain-mixed representation learning model is proposed accordingly to encode the multiple subdomains effectively. The whole process is performed progressively for both corpus construction and model training. Experimental results on a benchmark dataset show that our method can gain significant improvements over a vary of baselines. Extensive analyses are performed to show the advantages of our final domain adaption model as well.

Relation detection in knowledge base question answering, aims to identify the path(s) of relations starting from the topic entity node that is linked to the answer node in knowledge graph. Such path might consist of multiple relations, which we call multi-hop. Moreover, for a single question, there may exist multiple relation paths to the correct answer, which we call multi-label. However, most of existing approaches only detect one single path to obtain the answer without considering other correct paths, which might affect the final performance. Therefore, in this paper, we propose a novel divide-and-conquer approach for multi-label multi-hop relation detection (DC-MLMH) by decomposing it into head relation detection and conditional relation path generation. In specific, a novel path sampling mechanism is proposed to generate diverse relation paths for the inference stage. A majority-vote policy is employed to detect final KB answer. Comprehensive experiments were conducted on the FreebaseQA benchmark dataset. Experimental results show that the proposed approach not only outperforms other competitive multi-label baselines, but also has superiority over some state-of-art KBQA methods.

Emotion recognition in multi-party conversation (ERMC) is becoming increasingly popular as an emerging research topic in natural language processing. Prior research focuses on exploring sequential information but ignores the discourse structures of conversations. In this paper, we investigate the importance of discourse structures in handling informative contextual cues and speaker-specific features for ERMC. To this end, we propose a discourse-aware graph neural network (ERMC-DisGCN) for ERMC. In particular, we design a relational convolution to lever the self-speaker dependency of interlocutors to propagate contextual information. Furthermore, we exploit a gated convolution to select more informative cues for ERMC from dependent utterances. The experimental results show our method outperforms multiple baselines, illustrating that discourse structures are of great value to ERMC.