COLING 2022 论文列表

In natural language, referencing objects at different levels of specificity is a fundamental pragmatic mechanism for efficient communication in context. We develop a novel communication game, the hierarchical reference game, to study the emergence of such reference systems in artificial agents. We consider a simplified world, in which concepts are abstractions over a set of primitive attributes (e.g., color, style, shape). Depending on how many attributes are combined, concepts are more general (“circle”) or more specific (“red dotted circle”). Based on the context, the agents have to communicate at different levels of this hierarchy. Our results show that the agents learn to play the game successfully and can even generalize to novel concepts. To achieve abstraction, they use implicit (omitting irrelevant information) and explicit (indicating that attributes are irrelevant) strategies. In addition, the compositional structure underlying the concept hierarchy is reflected in the emergent protocols, indicating that the need to develop hierarchical reference systems supports the emergence of compositionality.

Learning visual and textual representations in the shared space from web-scale image-text pairs improves the performance of diverse vision-and-language tasks, as well as modality-specific tasks. Many attempts in this framework have been made to connect English-only texts and images, and only a few works have been proposed to extend this framework in multilingual settings with the help of many translation pairs. In this multilingual approach, a typical setup is to use pairs of (image and English-text) and translation pairs. The major limitation of this approach is that the learning signal of aligning visual representation with under-resourced language representation is not strong, achieving a sub-optimal performance of vision-and-language tasks. In this work, we propose a simple yet effective enhancement scheme for previous multilingual multi-modal representation methods by using a limited number of pairs of images and non-English texts. In specific, our scheme fine-tunes a pre-trained multilingual model by minimizing a triplet contrastive loss on triplets of image and two different language texts with the same meaning, improving the connection between images and non-English texts. Experiments confirm that our enhancement strategy achieves performance gains in image-text retrieval, zero-shot image classification, and sentence embedding tasks.

In recent years, data augmentation has become an important field of machine learning. While images can use simple techniques such as cropping or rotating, textual data augmentation needs more complex manipulations to ensure that the generated examples are useful. Variational auto-encoders (VAE) and its conditional variant the Conditional-VAE (CVAE) are often used to generate new textual data, both relying on a good enough training of the generator so that it doesn’t create examples of the wrong class. In this paper, we explore a simpler way to use VAE for data augmentation: the training of one VAE per class. We show on several dataset sizes, as well as on four different binary classification tasks, that it systematically outperforms other generative data augmentation techniques.

Post-processing of static embedding has been shown to improve their performance on both lexical and sequence-level tasks. However, post-processing for contextualized embeddings is an under-studied problem. In this work, we question the usefulness of post-processing for contextualized embeddings obtained from different layers of pre-trained language models. More specifically, we standardize individual neuron activations using z-score, min-max normalization, and by removing top principal components using the all-but-the-top method. Additionally, we apply unit length normalization to word representations. On a diverse set of pre-trained models, we show that post-processing unwraps vital information present in the representations for both lexical tasks (such as word similarity and analogy) and sequence classification tasks. Our findings raise interesting points in relation to the research studies that use contextualized representations, and suggest z-score normalization as an essential step to consider when using them in an application.

Conversational machine reading comprehension (CMRC) aims to assist computers to understand an natural language text and thereafter engage in a multi-turn conversation to answer questions related to the text. Existing methods typically require three steps: (1) decision making based on entailment reasoning; (2) span extraction if required by the above decision; (3) question rephrasing based on the extracted span. However, for nearly all these methods, the span extraction and question rephrasing steps cannot fully exploit the fine-grained entailment reasoning information in decision making step because of their relative independence, which will further enlarge the information gap between decision making and question phrasing. Thus, to tackle this problem, we propose a novel end-to-end framework for conversational machine reading comprehension based on shared parameter mechanism, called entailment reasoning T5 (ET5). Despite the lightweight of our proposed framework, experimental results show that the proposed ET5 achieves new state-of-the-art results on the ShARC leaderboard with the BLEU-4 score of 55.2. Our model and code are publicly available.

Contrastive learning has been attracting much attention for learning unsupervised sentence embeddings. The current state-of-the-art unsupervised method is the unsupervised SimCSE (unsup-SimCSE). Unsup-SimCSE takes dropout as a minimal data augmentation method, and passes the same input sentence to a pre-trained Transformer encoder (with dropout turned on) twice to obtain the two corresponding embeddings to build a positive pair. As the length information of a sentence will generally be encoded into the sentence embeddings due to the usage of position embedding in Transformer, each positive pair in unsup-SimCSE actually contains the same length information. And thus unsup-SimCSE trained with these positive pairs is probably biased, which would tend to consider that sentences of the same or similar length are more similar in semantics. Through statistical observations, we find that unsup-SimCSE does have such a problem. To alleviate it, we apply a simple repetition operation to modify the input sentence, and then pass the input sentence and its modified counterpart to the pre-trained Transformer encoder, respectively, to get the positive pair. Additionally, we draw inspiration from the community of computer vision and introduce a momentum contrast, enlarging the number of negative pairs without additional calculations. The proposed two modifications are applied on positive and negative pairs separately, and build a new sentence embedding method, termed Enhanced Unsup-SimCSE (ESimCSE). We evaluate the proposed ESimCSE on several benchmark datasets w.r.t the semantic text similarity (STS) task. Experimental results show that ESimCSE outperforms the state-of-the-art unsup-SimCSE by an average Spearman correlation of 2.02% on BERT-base.

Document-level Event Causality Identification (DECI) aims to identify event-event causal relations in a document. Existing works usually build an event graph for global reasoning across multiple sentences. However, the edges between events have to be carefully designed through heuristic rules or external tools. In this paper, we propose a novel Event Relational Graph TransfOrmer (ERGO) framework for DECI, to ease the graph construction and improve it over the noisy edge issue. Different from conventional event graphs, we define a pair of events as a node and build a complete event relational graph without any prior knowledge or tools. This naturally formulates DECI as a node classification problem, and thus we capture the causation transitivity among event pairs via a graph transformer. Furthermore, we design a criss-cross constraint and an adaptive focal loss for the imbalanced classification, to alleviate the issues of false positives and false negatives. Extensive experiments on two benchmark datasets show that ERGO greatly outperforms previous state-of-the-art (SOTA) methods (12.8% F1 gains on average).

The World Health Organization has emphasised the need of stepping up suicide prevention efforts to meet the United Nation’s Sustainable Development Goal target of 2030 (Goal 3: Good health and well-being). We address the challenging task of personality subtyping from suicide notes. Most research on personality subtyping has relied on statistical analysis and feature engineering. Moreover, state-of-the-art transformer models in the automated personality subtyping problem have received relatively less attention. We develop a novel EMotion-assisted PERSONAlity Detection Framework (EM-PERSONA). We annotate the benchmark CEASE-v2.0 suicide notes dataset with personality traits across four dichotomies: Introversion (I)-Extraversion (E), Intuition (N)-Sensing (S), Thinking (T)-Feeling (F), Judging (J)–Perceiving (P). Our proposed method outperforms all baselines on comprehensive evaluation using multiple state-of-the-art systems. Across the four dichotomies, EM-PERSONA improved accuracy by 2.04%, 3.69%, 4.52%, and 3.42%, respectively, over the highest-performing single-task systems.

Enhancing the interpretability of text classification models can help increase the reliability of these models in real-world applications. Currently, most researchers focus on extracting task-specific words from inputs to improve the interpretability of the model. The competitive approaches exploit the Variational Information Bottleneck (VIB) to improve the performance of word masking at the word embedding layer to obtain task-specific words. However, these approaches ignore the multi-level semantics of the text, which can impair the interpretability of the model, and do not consider the risk of representation overlap caused by the VIB, which can impair the classification performance. In this paper, we propose an enhanced variational word masks approach, named E-VarM, to solve these two issues effectively. The E-VarM combines multi-level semantics from all hidden layers of the model to mask out task-irrelevant words and uses contrastive learning to readjust the distances between representations. Empirical studies on ten benchmark text classification datasets demonstrate that our approach outperforms the SOTA methods in simultaneously improving the interpretability and accuracy of the model.

This work investigates the challenge of learning and reasoning for Commonsense Question Answering given an external source of knowledge in the form of a knowledge graph (KG). We propose a novel graph neural network architecture, called Dynamic Relevance Graph Network (DRGN). DRGN operates on a given KG subgraph based on the question and answers entities and uses the relevance scores between the nodes to establish new edges dynamically for learning node representations in the graph network. This explicit usage of relevance as graph edges has the following advantages, a) the model can exploit the existing relationships, re-scale the node weights, and influence the way the neighborhood nodes’ representations are aggregated in the KG subgraph, b) It potentially recovers the missing edges in KG that are needed for reasoning. Moreover, as a byproduct, our model improves handling the negative questions due to considering the relevance between the question node and the graph entities. Our proposed approach shows competitive performance on two QA benchmarks, CommonsenseQA and OpenbookQA, compared to the state-of-the-art published results.

Recurrent models have been dominating the field of neural machine translation (NMT) for the past few years. Transformers have radically changed it by proposing a novel architecture that relies on a feed-forward backbone and self-attention mechanism. Although Transformers are powerful, they could fail to properly encode sequential/positional information due to their non-recurrent nature. To solve this problem, position embeddings are defined exclusively for each time step to enrich word information. However, such embeddings are fixed after training regardless of the task and word ordering system of the source and target languages. In this paper, we address this shortcoming by proposing a novel architecture with new position embeddings that take the order of the target words into consideration. Instead of using predefined position embeddings, our solution generates new embeddings to refine each word’s position information. Since we do not dictate the position of the source tokens and we learn them in an end-to-end fashion, we refer to our method as dynamic position encoding (DPE). We evaluated the impact of our model on multiple datasets to translate from English to German, French, and Italian and observed meaningful improvements in comparison to the original Transformer.

Data augmentation with mixup has shown to be effective on the NLP tasks. Although its great success, the mixup still has shortcomings. First, vanilla mixup randomly selects one sample to generate the mixup sample for a given sample. It remains unclear how to best choose the input samples for the mixup. Second, linear interpolation limits the space of synthetic data and its regularization effect. In this paper, we propose the dynamic nonlinear mixup with distance-based sample selection, which not only generates multiple sample pairs based on the distance between each sample but also enlarges the space of synthetic samples. Specifically, we compute the distance between each input data by cosine similarity and select multiple samples for a given sample. Then we use the dynamic nonlinear mixup to fuse sample pairs. It does not use a linear, scalar mixing strategy, but a nonlinear interpolation strategy, where the mixing strategy is adaptively updated for the input and label pairs. Experiments on the multiple public datasets demonstrate that dynamic nonlinear mixup outperforms state-of-the-art methods.

Continual learning is one of the key components of human learning and a necessary requirement of artificial intelligence. As dialogue can potentially span infinitely many topics and tasks, a task-oriented dialogue system must have the capability to continually learn, dynamically adapting to new challenges while preserving the knowledge it already acquired. Despite the importance, continual reinforcement learning of the dialogue policy has remained largely unaddressed. The lack of a framework with training protocols, baseline models and suitable metrics, has so far hindered research in this direction. In this work we fill precisely this gap, enabling research in dialogue policy optimisation to go from static to dynamic learning. We provide a continual learning algorithm, baseline architectures and metrics for assessing continual learning models. Moreover, we propose the dynamic dialogue policy transformer (DDPT), a novel dynamic architecture that can integrate new knowledge seamlessly, is capable of handling large state spaces and obtains significant zero-shot performance when being exposed to unseen domains, without any growth in network parameter size. We validate the strengths of DDPT in simulation with two user simulators as well as with humans.

A recent success in semantic dependency parsing shows that graph neural networks can make significant accuracy improvements, owing to its powerful ability in learning expressive graph representations. However, this work learns graph representations based on a static graph constructed by an existing parser, suffering from two drawbacks: (1) the static graph might be error-prone (e.g., noisy or incomplete), and (2) graph construction stage and graph representation learning stage are disjoint, the errors introduced in the graph construction stage cannot be corrected and might be accumulated to later stages. To address these two drawbacks, we propose a dynamic graph learning framework and apply it to semantic dependency parsing, for jointly learning graph structure and graph representations. Experimental results show that our parser outperforms the previous parsers on the SemEval-2015 Task 18 dataset in three languages (English, Chinese, and Czech).

A Visual Question Answering (VQA) model processes images and questions simultaneously with rich semantic information. The attention mechanism can highlight fine-grained features with critical information, thus ensuring that feature extraction emphasizes the objects related to the questions. However, unattended coarse-grained information is also essential for questions involving global elements. We believe that global coarse-grained information and local fine-grained information can complement each other to provide richer comprehensive information. In this paper, we propose a dual capsule attention mask network with mutual learning for VQA. Specifically, it contains two branches processing coarse-grained features and fine-grained features, respectively. We also design a novel stackable dual capsule attention module to fuse features and locate evidence. The two branches are combined to make final predictions for VQA. Experimental results show that our method outperforms the baselines in terms of VQA performance and interpretability and achieves new SOTA performance on the VQA-v2 dataset.

This paper proposes a simple yet effective interpolation-based data augmentation approach termed DoubleMix, to improve the robustness of models in text classification. DoubleMix first leverages a couple of simple augmentation operations to generate several perturbed samples for each training data, and then uses the perturbed data and original data to carry out a two-step interpolation in the hidden space of neural models. Concretely, it first mixes up the perturbed data to a synthetic sample and then mixes up the original data and the synthetic perturbed data. DoubleMix enhances models’ robustness by learning the “shifted” features in hidden space. On six text classification benchmark datasets, our approach outperforms several popular text augmentation methods including token-level, sentence-level, and hidden-level data augmentation techniques. Also, experiments in low-resource settings show our approach consistently improves models’ performance when the training data is scarce. Extensive ablation studies and case studies confirm that each component of our approach contributes to the final performance and show that our approach exhibits superior performance on challenging counterexamples. Additionally, visual analysis shows that text features generated by our approach are highly interpretable.

Recent pre-trained language models (PLMs) achieved great success on many natural language processing tasks through learning linguistic features and contextualized sentence representation. Since attributes captured in stacked layers of PLMs are not clearly identified, straightforward approaches such as embedding the last layer are commonly preferred to derive sentence representations from PLMs. This paper introduces the attention-based pooling strategy, which enables the model to preserve layer-wise signals captured in each layer and learn digested linguistic features for downstream tasks. The contrastive learning objective can adapt the layer-wise attention pooling to both unsupervised and supervised manners. It results in regularizing the anisotropic space of pre-trained embeddings and being more uniform. We evaluate our model on standard semantic textual similarity (STS) and semantic search tasks. As a result, our method improved the performance of the base contrastive learned BERTbase and variants.

Successful Machine Learning based Named Entity Recognition models could fail on texts from some special domains, for instance, Chinese addresses and e-commerce titles, where requires adequate background knowledge. Such texts are also difficult for human annotators. In fact, we can obtain some potentially helpful information from correlated texts, which have some common entities, to help the text understanding. Then, one can easily reason out the correct answer by referencing correlated samples. In this paper, we suggest enhancing NER models with correlated samples. We draw correlated samples by the sparse BM25 retriever from large-scale in-domain unlabeled data. To explicitly simulate the human reasoning process, we perform a training-free entity type calibrating by majority voting. To capture correlation features in the training stage, we suggest to model correlated samples by the transformer-based multi-instance cross-encoder. Empirical results on datasets of the above two domains show the efficacy of our methods.

FAQs are important resources to find information. However, especially if a FAQ concerns many question-answer pairs, it can be a difficult and time-consuming job to find the answer you are looking for. A FAQ chatbot can ease this process by automatically retrieving the relevant answer to a user’s question. We present VaccinChatNL, a Dutch FAQ corpus on the topic of COVID-19 vaccination. Starting with 50 question-answer pairs we built VaccinChat, a FAQ chatbot, which we used to gather more user questions that were also annotated with the appropriate or new answer classes. This iterative process of gathering user questions, annotating them, and retraining the model with the increased data set led to a corpus that now contains 12,883 user questions divided over 181 answers. We provide the first publicly available Dutch FAQ answering data set of this size with large groups of semantically equivalent human-paraphrased questions. Furthermore, our study shows that before fine-tuning a classifier, continued pre-training of Dutch language models with task- and/or domain-specific data improves classification results. In addition, we show that large groups of semantically similar questions are important for obtaining well-performing intent classification models.