EMNLP 2021 Papers

Neural dialog models are known to suffer from problems such as generating unsafe and inconsistent responses. Even though these problems are crucial and prevalent, they are mostly manually identified by model designers through interactions. Recently, some research instructs crowdworkers to goad the bots into triggering such problems. However, humans leverage superficial clues such as hate speech, while leaving systematic problems undercover. In this paper, we propose two methods including reinforcement learning to automatically trigger a dialog model into generating problematic responses. We show the effect of our methods in exposing safety and contradiction issues with state-of-the-art dialog models.

The understanding of time expressions includes two sub-tasks: recognition and normalization. In recent years, significant progress has been made in the recognition of time expressions while research on normalization has lagged behind. Existing SOTA normalization methods highly rely on rules or grammars designed by experts, which limits their performance on emerging corpora, such as social media texts. In this paper, we model time expression normalization as a sequence of operations to construct the normalized temporal value, and we present a novel method called ARTime, which can automatically generate normalization rules from training data without expert interventions. Specifically, ARTime automatically captures possible operation sequences from annotated data and generates normalization rules on time expressions with common surface forms. The experimental results show that ARTime can significantly surpass SOTA methods on the Tweets benchmark, and achieves competitive results with existing expert-engineered rule methods on the TempEval-3 benchmark.

A new metric BaryScore to evaluate text generation based on deep contextualized embeddings (e.g., BERT, Roberta, ELMo) is introduced. This metric is motivated by a new framework relying on optimal transport tools, i.e., Wasserstein distance and barycenter. By modelling the layer output of deep contextualized embeddings as a probability distribution rather than by a vector embedding; this framework provides a natural way to aggregate the different outputs through the Wasserstein space topology. In addition, it provides theoretical grounds to our metric and offers an alternative to available solutions (e.g., MoverScore and BertScore). Numerical evaluation is performed on four different tasks: machine translation, summarization, data2text generation and image captioning. Our results show that BaryScore outperforms other BERT based metrics and exhibits more consistent behaviour in particular for text summarization.

In this paper, we address the problem of automatically discriminating between inherited and borrowed Latin words. We introduce a new dataset and investigate the case of Romance languages (Romanian, Italian, French, Spanish, Portuguese and Catalan), where words directly inherited from Latin coexist with words borrowed from Latin, and explore whether automatic discrimination between them is possible. Having entered the language at a later stage, borrowed words are no longer subject to historical sound shift rules, hence they are presumably less eroded, which is why we expect them to have a different intrinsic structure distinguishable by computational means. We employ several machine learning models to automatically discriminate between inherited and borrowed words and compare their performance with various feature sets. We analyze the models’ predictive power on two versions of the datasets, orthographic and phonetic. We also investigate whether prior knowledge of the etymon provides better results, employing n-gram character features extracted from the word-etymon pairs and from their alignment.

In this paper, we present an automatic knowledge base construction system from large scale enterprise documents with minimal efforts of human intervention. In the design and deployment of such a knowledge mining system for enterprise, we faced several challenges including data distributional shift, performance evaluation, compliance requirements and other practical issues. We leveraged state-of-the-art deep learning models to extract information (named entities and definitions) at per document level, then further applied classical machine learning techniques to process global statistical information to improve the knowledge base. Experimental results are reported on actual enterprise documents. This system is currently serving as part of a Microsoft 365 service.

We describe the task of bilingual markup transfer, which involves placing markup tags from a source sentence into a fixed target translation. This task arises in practice when a human translator generates the target translation without markup, and then the system infers the placement of markup tags. This task contrasts from previous work in which markup transfer is performed jointly with machine translation. We propose two novel metrics and evaluate several approaches based on unsupervised word alignments as well as a supervised neural sequence-to-sequence model. Our best approach achieves an average accuracy of 94.7% across six language pairs, indicating its potential usefulness for real-world localization tasks.

Our paper aims to automate the generation of medical reports from chest X-ray image inputs, a critical yet time-consuming task for radiologists. Existing medical report generation efforts emphasize producing human-readable reports, yet the generated text may not be well aligned to the clinical facts. Our generated medical reports, on the other hand, are fluent and, more importantly, clinically accurate. This is achieved by our fully differentiable and end-to-end paradigm that contains three complementary modules: taking the chest X-ray images and clinical history document of patients as inputs, our classification module produces an internal checklist of disease-related topics, referred to as enriched disease embedding; the embedding representation is then passed to our transformer-based generator, to produce the medical report; meanwhile, our generator also creates a weighted embedding representation, which is fed to our interpreter to ensure consistency with respect to disease-related topics. Empirical evaluations demonstrate very promising results achieved by our approach on commonly-used metrics concerning language fluency and clinical accuracy. Moreover, noticeable performance gains are consistently observed when additional input information is available, such as the clinical document and extra scans from different views.

There has been a significant progress in the field of Extractive Question Answering (EQA) in the recent years. However, most of them are reliant on annotations of answer-spans in the corresponding passages. In this work, we address the problem of EQA when no annotations are present for the answer span, i.e., when the dataset contains only questions and corresponding passages. Our method is based on auto-encoding of the question that performs a question answering task during encoding and a question generation task during decoding. We show that our method performs well in a zero-shot setting and can provide an additional loss to boost performance for EQA.

Current language models are usually trained using a self-supervised scheme, where the main focus is learning representations at the word or sentence level. However, there has been limited progress in generating useful discourse-level representations. In this work, we propose to use ideas from predictive coding theory to augment BERT-style language models with a mechanism that allows them to learn suitable discourse-level representations. As a result, our proposed approach is able to predict future sentences using explicit top-down connections that operate at the intermediate layers of the network. By experimenting with benchmarks designed to evaluate discourse-related knowledge using pre-trained sentence representations, we demonstrate that our approach improves performance in 6 out of 11 tasks by excelling in discourse relationship detection.

Large language models benefit from training with a large amount of unlabeled text, which gives them increasingly fluent and diverse generation capabilities. However, using these models for text generation that takes into account target attributes, such as sentiment polarity or specific topics, remains a challenge. We propose a simple and flexible method for controlling text generation by aligning disentangled attribute representations. In contrast to recent efforts on training a discriminator to perturb the token level distribution for an attribute, we use the same data to learn an alignment function to guide the pre-trained, non-controlled language model to generate texts with the target attribute without changing the original language model parameters. We evaluate our method on sentiment- and topic-controlled generation, and show large performance gains over previous methods while retaining fluency and diversity.

Keyword or keyphrase extraction is to identify words or phrases presenting the main topics of a document. This paper proposes the AttentionRank, a hybrid attention model, to identify keyphrases from a document in an unsupervised manner. AttentionRank calculates self-attention and cross-attention using a pre-trained language model. The self-attention is designed to determine the importance of a candidate within the context of a sentence. The cross-attention is calculated to identify the semantic relevance between a candidate and sentences within a document. We evaluate the AttentionRank on three publicly available datasets against seven baselines. The results show that the AttentionRank is an effective and robust unsupervised keyphrase extraction model on both long and short documents. Source code is available on Github.

Self-supervised learning has recently attracted considerable attention in the NLP community for its ability to learn discriminative features using a contrastive objective. This paper investigates whether contrastive learning can be extended to Transfomer attention to tackling the Winograd Schema Challenge. To this end, we propose a novel self-supervised framework, leveraging a contrastive loss directly at the level of self-attention. Experimental analysis of our attention-based models on multiple datasets demonstrates superior commonsense reasoning capabilities. The proposed approach outperforms all comparable unsupervised approaches while occasionally surpassing supervised ones.

This work proposes an extensive analysis of the Transformer architecture in the Neural Machine Translation (NMT) setting. Focusing on the encoder-decoder attention mechanism, we prove that attention weights systematically make alignment errors by relying mainly on uninformative tokens from the source sequence. However, we observe that NMT models assign attention to these tokens to regulate the contribution in the prediction of the two contexts, the source and the prefix of the target sequence. We provide evidence about the influence of wrong alignments on the model behavior, demonstrating that the encoder-decoder attention mechanism is well suited as an interpretability method for NMT. Finally, based on our analysis, we propose methods that largely reduce the word alignment error rate compared to standard induced alignments from attention weights.

Few-shot table-to-text generation is a task of composing fluent and faithful sentences to convey table content using limited data. Despite many efforts having been made towards generating impressive fluent sentences by fine-tuning powerful pre-trained language models, the faithfulness of generated content still needs to be improved. To this end, this paper proposes a novel approach Attend, Memorize and Generate (called AMG), inspired by the text generation process of humans. In particular, AMG (1) attends over the multi-granularity of context using a novel strategy based on table slot level and traditional token-by-token level attention to exploit both the table structure and natural linguistic information; (2) dynamically memorizes the table slot allocation states; and (3) generates faithful sentences according to both the context and memory allocation states. Comprehensive experiments with human evaluation on three domains (i.e., humans, songs, and books) of the Wiki dataset show that our model can generate higher qualified texts when compared with several state-of-the-art baselines, in both fluency and faithfulness.

Athena 2.0 is an Alexa Prize SocialBot that has been a finalist in the last two Alexa Prize Grand Challenges. One reason for Athena’s success is its novel dialogue management strategy, which allows it to dynamically construct dialogues and responses from component modules, leading to novel conversations with every interaction. Here we describe Athena’s system design and performance in the Alexa Prize during the 20/21 competition. A live demo of Athena as well as video recordings will provoke discussion on the state of the art in conversational AI.

Various measures have been proposed to quantify human-like social biases in word embeddings. However, bias scores based on these measures can suffer from measurement error. One indication of measurement quality is reliability, concerning the extent to which a measure produces consistent results. In this paper, we assess three types of reliability of word embedding gender bias measures, namely test-retest reliability, inter-rater consistency and internal consistency. Specifically, we investigate the consistency of bias scores across different choices of random seeds, scoring rules and words. Furthermore, we analyse the effects of various factors on these measures’ reliability scores. Our findings inform better design of word embedding gender bias measures. Moreover, we urge researchers to be more critical about the application of such measures

Aspect-based sentiment analysis (ABSA) typically focuses on extracting aspects and predicting their sentiments on individual sentences such as customer reviews. Recently, another kind of opinion sharing platform, namely question answering (QA) forum, has received increasing popularity, which accumulates a large number of user opinions towards various aspects. This motivates us to investigate the task of ABSA on QA forums (ABSA-QA), aiming to jointly detect the discussed aspects and their sentiment polarities for a given QA pair. Unlike review sentences, a QA pair is composed of two parallel sentences, which requires interaction modeling to align the aspect mentioned in the question and the associated opinion clues in the answer. To this end, we propose a model with a specific design of cross-sentence aspect-opinion interaction modeling to address this task. The proposed method is evaluated on three real-world datasets and the results show that our model outperforms several strong baselines adopted from related state-of-the-art models.

Recent work on opinion summarization produces general summaries based on a set of input reviews and the popularity of opinions expressed in them. In this paper, we propose an approach that allows the generation of customized summaries based on aspect queries (e.g., describing the location and room of a hotel). Using a review corpus, we create a synthetic training dataset of (review, summary) pairs enriched with aspect controllers which are induced by a multi-instance learning model that predicts the aspects of a document at different levels of granularity. We fine-tune a pretrained model using our synthetic dataset and generate aspect-specific summaries by modifying the aspect controllers. Experiments on two benchmarks show that our model outperforms the previous state of the art and generates personalized summaries by controlling the number of aspects discussed in them.

Aspect-based sentiment analysis (ABSA) has been extensively studied in recent years, which typically involves four fundamental sentiment elements, including the aspect category, aspect term, opinion term, and sentiment polarity. Existing studies usually consider the detection of partial sentiment elements, instead of predicting the four elements in one shot. In this work, we introduce the Aspect Sentiment Quad Prediction (ASQP) task, aiming to jointly detect all sentiment elements in quads for a given opinionated sentence, which can reveal a more comprehensive and complete aspect-level sentiment structure. We further propose a novel Paraphrase modeling paradigm to cast the ASQP task to a paraphrase generation process. On one hand, the generation formulation allows solving ASQP in an end-to-end manner, alleviating the potential error propagation in the pipeline solution. On the other hand, the semantics of the sentiment elements can be fully exploited by learning to generate them in the natural language form. Extensive experiments on benchmark datasets show the superiority of our proposed method and the capacity of cross-task transfer with the proposed unified Paraphrase modeling framework.