ACL 2022 论文列表

In recent years, machine learning models have rapidly become better at generating clinical consultation notes; yet, there is little work on how to properly evaluate the generated consultation notes to understand the impact they may have on both the clinician using them and the patient’s clinical safety. To address this we present an extensive human evaluation study of consultation notes where 5 clinicians (i) listen to 57 mock consultations, (ii) write their own notes, (iii) post-edit a number of automatically generated notes, and (iv) extract all the errors, both quantitative and qualitative. We then carry out a correlation study with 18 automatic quality metrics and the human judgements. We find that a simple, character-based Levenshtein distance metric performs on par if not better than common model-based metrics like BertScore. All our findings and annotations are open-sourced.

As privacy gains traction in the NLP community, researchers have started adopting various approaches to privacy-preserving methods. One of the favorite privacy frameworks, differential privacy (DP), is perhaps the most compelling thanks to its fundamental theoretical guarantees. Despite the apparent simplicity of the general concept of differential privacy, it seems non-trivial to get it right when applying it to NLP. In this short paper, we formally analyze several recent NLP papers proposing text representation learning using DPText (Beigi et al., 2019a,b; Alnasser et al., 2021; Beigi et al., 2021) and reveal their false claims of being differentially private. Furthermore, we also show a simple yet general empirical sanity check to determine whether a given implementation of a DP mechanism almost certainly violates the privacy loss guarantees. Our main goal is to raise awareness and help the community understand potential pitfalls of applying differential privacy to text representation learning.

Several pre-training objectives, such as masked language modeling (MLM), have been proposed to pre-train language models (e.g. BERT) with the aim of learning better language representations. However, to the best of our knowledge, no previous work so far has investigated how different pre-training objectives affect what BERT learns about linguistics properties. We hypothesize that linguistically motivated objectives such as MLM should help BERT to acquire better linguistic knowledge compared to other non-linguistically motivated objectives that are not intuitive or hard for humans to guess the association between the input and the label to be predicted. To this end, we pre-train BERT with two linguistically motivated objectives and three non-linguistically motivated ones. We then probe for linguistic characteristics encoded in the representation of the resulting models. We find strong evidence that there are only small differences in probing performance between the representations learned by the two different types of objectives. These surprising results question the dominant narrative of linguistically informed pre-training.

More than 43% of the languages spoken in the world are endangered, and language loss currently occurs at an accelerated rate because of globalization and neocolonialism. Saving and revitalizing endangered languages has become very important for maintaining the cultural diversity on our planet. In this work, we focus on discussing how NLP can help revitalize endangered languages. We first suggest three principles that may help NLP practitioners to foster mutual understanding and collaboration with language communities, and we discuss three ways in which NLP can potentially assist in language education. We then take Cherokee, a severely-endangered Native American language, as a case study. After reviewing the language’s history, linguistic features, and existing resources, we (in collaboration with Cherokee community members) arrive at a few meaningful ways NLP practitioners can collaborate with community partners. We suggest two approaches to enrich the Cherokee language’s resources with machine-in-the-loop processing, and discuss several NLP tools that people from the Cherokee community have shown interest in. We hope that our work serves not only to inform the NLP community about Cherokee, but also to provide inspiration for future work on endangered languages in general.

Recently, there has been a trend to investigate the factual knowledge captured by Pre-trained Language Models (PLMs). Many works show the PLMs’ ability to fill in the missing factual words in cloze-style prompts such as ”Dante was born in [MASK].” However, it is still a mystery how PLMs generate the results correctly: relying on effective clues or shortcut patterns? We try to answer this question by a causal-inspired analysis that quantitatively measures and evaluates the word-level patterns that PLMs depend on to generate the missing words. We check the words that have three typical associations with the missing words: knowledge-dependent, positionally close, and highly co-occurred. Our analysis shows: (1) PLMs generate the missing factual words more by the positionally close and highly co-occurred words than the knowledge-dependent words; (2) the dependence on the knowledge-dependent words is more effective than the positionally close and highly co-occurred words. Accordingly, we conclude that the PLMs capture the factual knowledge ineffectively because of depending on the inadequate associations.

Long-form answers, consisting of multiple sentences, can provide nuanced and comprehensive answers to a broader set of questions. To better understand this complex and understudied task, we study the functional structure of long-form answers collected from three datasets, ELI5, WebGPT and Natural Questions. Our main goal is to understand how humans organize information to craft complex answers. We develop an ontology of six sentence-level functional roles for long-form answers, and annotate 3.9k sentences in 640 answer paragraphs. Different answer collection methods manifest in different discourse structures. We further analyze model-generated answers – finding that annotators agree less with each other when annotating model-generated answers compared to annotating human-written answers. Our annotated data enables training a strong classifier that can be used for automatic analysis. We hope our work can inspire future research on discourse-level modeling and evaluation of long-form QA systems.

With the rapid development of deep learning, Seq2Seq paradigm has become prevalent for end-to-end data-to-text generation, and the BLEU scores have been increasing in recent years. However, it is widely recognized that there is still a gap between the quality of the texts generated by models and the texts written by human. In order to better understand the ability of Seq2Seq models, evaluate their performance and analyze the results, we choose to use Multidimensional Quality Metric(MQM) to evaluate several representative Seq2Seq models on end-to-end data-to-text generation. We annotate the outputs of five models on four datasets with eight error types and find that 1) copy mechanism is helpful for the improvement in Omission and Inaccuracy Extrinsic errors but it increases other types of errors such as Addition; 2) pre-training techniques are highly effective, and pre-training strategy and model size are very significant; 3) the structure of the dataset also influences the model’s performance greatly; 4) some specific types of errors are generally challenging for seq2seq models.

This work addresses the question of the localization of syntactic information encoded in the transformers representations. We tackle this question from two perspectives, considering the object-past participle agreement in French, by identifying, first, in which part of the sentence and, second, in which part of the representation the syntactic information is encoded. The results of our experiments, using probing, causal analysis and feature selection method, show that syntactic information is encoded locally in a way consistent with the French grammar.

Cross-lingual Entity Typing (CLET) aims at improving the quality of entity type prediction by transferring semantic knowledge learned from rich-resourced languages to low-resourced languages. In this paper, by utilizing multilingual transfer learning via the mixture-of-experts approach, our model dynamically capture the relationship between target language and each source language, and effectively generalize to predict types of unseen entities in new languages. Extensive experiments on multi-lingual datasets show that our method significantly outperforms multiple baselines and can robustly handle negative transfer. We questioned the relationship between language similarity and the performance of CLET. A series of experiments refute the commonsense that the more source the better, and suggest the Similarity Hypothesis for CLET.

We propose a generative model of paraphrase generation, that encourages syntactic diversity by conditioning on an explicit syntactic sketch. We introduce Hierarchical Refinement Quantized Variational Autoencoders (HRQ-VAE), a method for learning decompositions of dense encodings as a sequence of discrete latent variables that make iterative refinements of increasing granularity. This hierarchy of codes is learned through end-to-end training, and represents fine-to-coarse grained information about the input. We use HRQ-VAE to encode the syntactic form of an input sentence as a path through the hierarchy, allowing us to more easily predict syntactic sketches at test time. Extensive experiments, including a human evaluation, confirm that HRQ-VAE learns a hierarchical representation of the input space, and generates paraphrases of higher quality than previous systems.

Large pretrained models enable transfer learning to low-resource domains for language generation tasks. However, previous end-to-end approaches do not account for the fact that some generation sub-tasks, specifically aggregation and lexicalisation, can benefit from transfer learning in different extents. To exploit these varying potentials for transfer learning, we propose a new hierarchical approach for few-shot and zero-shot generation. Our approach consists of a three-moduled jointly trained architecture: the first module independently lexicalises the distinct units of information in the input as sentence sub-units (e.g. phrases), the second module recurrently aggregates these sub-units to generate a unified intermediate output, while the third module subsequently post-edits it to generate a coherent and fluent final text. We perform extensive empirical analysis and ablation studies on few-shot and zero-shot settings across 4 datasets. Automatic and human evaluation shows that the proposed hierarchical approach is consistently capable of achieving state-of-the-art results when compared to previous work.

Pre-trained models have achieved excellent performance on the dialogue task. However, for the continual increase of online chit-chat scenarios, directly fine-tuning these models for each of the new tasks not only explodes the capacity of the dialogue system on the embedded devices but also causes knowledge forgetting on pre-trained models and knowledge interference among diverse dialogue tasks. In this work, we propose a hierarchical inductive transfer framework to learn and deploy the dialogue skills continually and efficiently. First, we introduce the adapter module into pre-trained models for learning new dialogue tasks. As the only trainable module, it is beneficial for the dialogue system on the embedded devices to acquire new dialogue skills with negligible additional parameters. Then, for alleviating knowledge interference between tasks yet benefiting the regularization between them, we further design hierarchical inductive transfer that enables new tasks to use general knowledge in the base adapter without being misled by diverse knowledge in task-specific adapters. Empirical evaluation and analysis indicate that our framework obtains comparable performance under deployment-friendly model capacity.

Abstract Meaning Representation (AMR) parsing aims to translate sentences to semantic representation with a hierarchical structure, and is recently empowered by pretrained sequence-to-sequence models. However, there exists a gap between their flat training objective (i.e., equally treats all output tokens) and the hierarchical AMR structure, which limits the model generalization. To bridge this gap, we propose a Hierarchical Curriculum Learning (HCL) framework with Structure-level (SC) and Instance-level Curricula (IC). SC switches progressively from core to detail AMR semantic elements while IC transits from structure-simple to -complex AMR instances during training. Through these two warming-up processes, HCL reduces the difficulty of learning complex structures, thus the flat model can better adapt to the AMR hierarchy. Extensive experiments on AMR2.0, AMR3.0, structure-complex and out-of-distribution situations verify the effectiveness of HCL.

Tables are often created with hierarchies, but existing works on table reasoning mainly focus on flat tables and neglect hierarchical tables. Hierarchical tables challenge numerical reasoning by complex hierarchical indexing, as well as implicit relationships of calculation and semantics. We present a new dataset, HiTab, to study question answering (QA) and natural language generation (NLG) over hierarchical tables. HiTab is a cross-domain dataset constructed from a wealth of statistical reports and Wikipedia pages, and has unique characteristics: (1) nearly all tables are hierarchical, and (2) QA pairs are not proposed by annotators from scratch, but are revised from real and meaningful sentences authored by analysts. (3) to reveal complex numerical reasoning in statistical reports, we provide fine-grained annotations of quantity and entity alignment. Experiments suggest that this HiTab presents a strong challenge for existing baselines and a valuable benchmark for future research. Targeting hierarchical structure, we devise a hierarchy-aware logical form for symbolic reasoning over tables, which shows high effectiveness. Targeting table reasoning, we leverage entity and quantity alignment to explore partially supervised training in QA and conditional generation in NLG, and largely reduce spurious predictions in QA and produce better descriptions in NLG.

Transformer-based language models usually treat texts as linear sequences. However, most texts also have an inherent hierarchical structure, i.e., parts of a text can be identified using their position in this hierarchy. In addition, section titles usually indicate the common topic of their respective sentences. We propose a novel approach to formulate, extract, encode and inject hierarchical structure information explicitly into an extractive summarization model based on a pre-trained, encoder-only Transformer language model (HiStruct+ model), which improves SOTA ROUGEs for extractive summarization on PubMed and arXiv substantially. Using various experimental settings on three datasets (i.e., CNN/DailyMail, PubMed and arXiv), our HiStruct+ model outperforms a strong baseline collectively, which differs from our model only in that the hierarchical structure information is not injected. It is also observed that the more conspicuous hierarchical structure the dataset has, the larger improvements our method gains. The ablation study demonstrates that the hierarchical position information is the main contributor to our model’s SOTA performance.

Distant supervision assumes that any sentence containing the same entity pairs reflects identical relationships. Previous works of distantly supervised relation extraction (DSRE) task generally focus on sentence-level or bag-level de-noising techniques independently, neglecting the explicit interaction with cross levels. In this paper, we propose a hierarchical contrastive learning Framework for Distantly Supervised relation extraction (HiCLRE) to reduce noisy sentences, which integrate the global structural information and local fine-grained interaction. Specifically, we propose a three-level hierarchical learning framework to interact with cross levels, generating the de-noising context-aware representations via adapting the existing multi-head self-attention, named Multi-Granularity Recontextualization. Meanwhile, pseudo positive samples are also provided in the specific level for contrastive learning via a dynamic gradient-based data augmentation strategy, named Dynamic Gradient Adversarial Perturbation. Experiments demonstrate that HiCLRE significantly outperforms strong baselines in various mainstream DSRE datasets.

Training giant models from scratch for each complex task is resource- and data-inefficient. To help develop models that can leverage existing systems, we propose a new challenge: Learning to solve complex tasks by communicating with existing agents (or models) in natural language. We design a synthetic benchmark, CommaQA, with three complex reasoning tasks (explicit, implicit, numeric) designed to be solved by communicating with existing QA agents. For instance, using text and table QA agents to answer questions such as “Who had the longest javelin throw from USA?”. We show that black-box models struggle to learn this task from scratch (accuracy under 50%) even with access to each agent’s knowledge and gold facts supervision. In contrast, models that learn to communicate with agents outperform black-box models, reaching scores of 100% when given gold decomposition supervision. However, we show that the challenge of learning to solve complex tasks by communicating with existing agents without relying on any auxiliary supervision or data still remains highly elusive. We will release CommaQA, along with a compositional generalization test split, to advance research in this direction.

Recently, various response generation models for two-party conversations have achieved impressive improvements, but less effort has been paid to multi-party conversations (MPCs) which are more practical and complicated. Compared with a two-party conversation where a dialogue context is a sequence of utterances, building a response generation model for MPCs is more challenging, since there exist complicated context structures and the generated responses heavily rely on both interlocutors (i.e., speaker and addressee) and history utterances. To address these challenges, we present HeterMPC, a heterogeneous graph-based neural network for response generation in MPCs which models the semantics of utterances and interlocutors simultaneously with two types of nodes in a graph. Besides, we also design six types of meta relations with node-edge-type-dependent parameters to characterize the heterogeneous interactions within the graph. Through multi-hop updating, HeterMPC can adequately utilize the structural knowledge of conversations for response generation. Experimental results on the Ubuntu Internet Relay Chat (IRC) channel benchmark show that HeterMPC outperforms various baseline models for response generation in MPCs.

We propose a new method for projective dependency parsing based on headed spans. In a projective dependency tree, the largest subtree rooted at each word covers a contiguous sequence (i.e., a span) in the surface order. We call such a span marked by a root word headed span. A projective dependency tree can be represented as a collection of headed spans. We decompose the score of a dependency tree into the scores of the headed spans and design a novel O(n3) dynamic programming algorithm to enable global training and exact inference. Our model achieves state-of-the-art or competitive results on PTB, CTB, and UD