EMNLP 2023 论文列表

Slot labeling (SL) is a core component of task-oriented dialogue (TOD) systems, where slots and corresponding values are usually language-, task- and domain-specific. Therefore, extending the system to any new language-domain-task configuration requires (re)running an expensive and resource-intensive data annotation process. To mitigate the inherent data scarcity issue, current research on multilingual ToD assumes that sufficient English-language annotated data are always available for particular tasks and domains, and thus operates in a standard cross-lingual transfer setup. In this work, we depart from this often unrealistic assumption. We examine challenging scenarios where such transfer-enabling English annotated data cannot be guaranteed, and focus on bootstrapping multilingual data-efficient slot labelers in transfer-free scenarios directly in the target languages without any English-ready data. We propose a two-stage slot labeling approach (termed TWOSL) which transforms standard multilingual sentence encoders into effective slot labelers. In Stage 1, relying on SL-adapted contrastive learning with only a handful of SL-annotated examples, we turn sentence encoders into task-specific span encoders. In Stage 2, we recast SL from a token classification into a simpler, less data-intensive span classification task. Our results on two standard multilingual TOD datasets and across diverse languages confirm the effectiveness and robustness of TWOSL. It is especially effective for the most challenging transfer-free few-shot setups, paving the way for quick and data-efficient bootstrapping of multilingual slot labelers for TOD.

Personalized dialogue systems aim to endow the chatbot agent with more anthropomorphic traits for human-like interactions. Previous approaches have explored explicitly user profile modeling using text descriptions, implicit derivation of user embeddings, or utilizing handicraft prompts for ChatGPT-like models. However, textual personas are limited in describing multi-faceted attributes (\emph{e.g.}, \emph{language style, inner character nuances}), implicit embedding suffers from personality sparsity, and handicraft prompts lack fine-grained and stable controllability. Hence, these approaches may struggle with complex personalized dialogue generation tasks that require generating controllable responses with multiple personal attributes. To this end, we propose \textbf{\textsc{Miracle}}, a novel personalized dialogue generation method through \textbf{M}ult\textbf{I}ple Pe\textbf{R}sonal \textbf{A}ttributes \textbf{C}ontrol within \textbf{L}atent-Space \textbf{E}nergy-based Models. ttributes \textbf{C}ontrol within \textbf{L}atent-Space \textbf{E}nergy-based Models. Specifically, our approach first disentangles complex personality into multi-faceted attributes. Subsequently, we employ a conditional variational auto-encoder to align with the dense personalized responses within a latent joint attribute space. We have also tailored a dedicated energy function and customized the ordinary differential equations sampling method to offer flexible attribute composition and precise attribute control. Extensive experiments demonstrate that \textsc{Miracle} outperforms several strong baselines in terms of personality controllability and response generation quality. Our dataset and code are available at \url{https://github.com/LZY-the-boys/MIRACLE}

Interlinear Morphological Glosses are annotations produced in the context of language documentation. Their goal is to identify morphs occurring in an L1 sentence and to explicit their function and meaning, with the further support of an associated translation in L2. We study here the task of automatic glossing, aiming to provide linguists with adequate tools to facilitate this process. Our formalisation of glossing uses a latent variable Conditional Random Field (CRF), which labels the L1 morphs while simultaneously aligning them to L2 words. In experiments with several under-resourced languages, we show that this approach is both effective and data-efficient and mitigates the problem of annotating unknown morphs. We also discuss various design choices regarding the alignment process and the selection of features. We finally demonstrate that it can benefit from multilingual (pre-)training, achieving results which outperform very strong baselines.

Unlike recurrent models, conventional wisdom has it that Transformers cannot perfectly model regular languages. Inspired by the notion of working memory, we propose a new Transformer variant named RegularGPT. With its novel combination of Weight-Sharing, Adaptive-Depth, and Sliding-Dilated-Attention, RegularGPT constructs working memory along the depth dimension, thereby enabling efficient and successful modeling of regular languages such as PARITY. We further test RegularGPT on the task of natural language length extrapolation and surprisingly find that it rediscovers the local windowed attention effect deemed necessary in prior work for length extrapolation.

Query rewriting plays a vital role in enhancing conversational search by transforming context-dependent user queries into standalone forms. Existing approaches primarily leverage human-rewritten queries as labels to train query rewriting models. However, human rewrites may lack sufficient information for optimal retrieval performance. To overcome this limitation, we propose utilizing large language models (LLMs) as query rewriters, enabling the generation of informative query rewrites through well-designed instructions. We define four essential properties for well-formed rewrites and incorporate all of them into the instruction. In addition, we introduce the role of rewrite editors for LLMs when initial query rewrites are available, forming a ``rewrite-then-edit'' process. Furthermore, we propose distilling the rewriting capabilities of LLMs into smaller models to reduce rewriting latency. Our experimental evaluation on the QReCC dataset demonstrates that informative query rewrites can yield substantially improved retrieval performance compared to human rewrites, especially with sparse retrievers.

The latest large language models (LLMs) such as ChatGPT, exhibit strong capabilities in automated mental health analysis. However, existing relevant studies bear several limitations, including inadequate evaluations, lack of prompting strategies, and ignorance of exploring LLMs for explainability. To bridge these gaps, we comprehensively evaluate the mental health analysis and emotional reasoning ability of LLMs on 11 datasets across 5 tasks. We explore the effects of different prompting strategies with unsupervised and distantly supervised emotional information. Based on these prompts, we explore LLMs for interpretable mental health analysis by instructing them to generate explanations for each of their decisions. We convey strict human evaluations to assess the quality of the generated explanations, leading to a novel dataset with 163 human-assessed explanations. We benchmark existing automatic evaluation metrics on this dataset to guide future related works. According to the results, ChatGPT shows strong in-context learning ability but still has a significant gap with advanced task-specific methods. Careful prompt engineering with emotional cues and expert-written few-shot examples can also effectively improve performance on mental health analysis. In addition, ChatGPT generates explanations that approach human performance, showing its great potential in explainable mental health analysis.

Providing explainable and faithful feedback is crucial for automated student answer assessment. In this paper, we introduce a novel framework that explores using ChatGPT, a cutting-edge large language model, for the concurrent tasks of student answer scoring and rationale generation. We identify the appropriate instructions by prompting ChatGPT with different templates to collect the rationales, where inconsistent rationales are refined to align with marking standards. The refined ChatGPT outputs enable us to fine-tune a smaller language model that simultaneously assesses student answers and provides rationales. Extensive experiments on the benchmark dataset show that the proposed method improves the overall QWK score by 11% compared to ChatGPT. Furthermore, our thorough analysis and human evaluation demonstrate that the rationales generated by our proposed method are comparable to those of ChatGPT. Our approach provides a viable solution to achieve explainable automated assessment in education

The task of Grammatical Error Correction (GEC) aims to automatically correct grammatical errors in natural texts. Almost all previous works treat annotated training data equally, but inherent discrepancies in data are neglected. In this paper, the inherent discrepancies are manifested in two aspects, namely, accuracy of data annotation and diversity of potential annotations. To this end, we propose MainGEC, which designs token-level and sentence-level training weights based on inherent discrepancies therein, and then conducts mixed-grained weighted training to improve the training effect for GEC. Empirical evaluation shows that whether in the Seq2Seq or Seq2Edit manner, MainGEC achieves consistent and significant performance improvements on two benchmark datasets, demonstrating the effectiveness and superiority of the mixed-grained weighted training. Further ablation experiments verify the effectiveness of designed weights for both granularities in MainGEC.

Synaesthesia refers to the description of perceptions in one sensory modality through concepts from other modalities. It involves not only a linguistic phenomenon, but also a cognitive phenomenon structuring human thought and action, which makes understanding it challenging. As a means of cognition, synaesthesia is rendered by more than sensory modalities, cue and stimulus can also play an important role in expressing and understanding it. In addition, understanding synaesthesia involves many cognitive efforts, such as identifying the semantic relationship between sensory words and modalities. Therefore, we propose a unified framework focusing on annotating all kinds of synaesthetic elements and fully exploring the relationship among them. In particular, we introduce a new annotation scheme, including sensory modalities as well as their cues and stimuli, which facilitate understanding synaesthetic information collectively. We further design a structure generation model to capture the relations among synaesthetic elements and generate them jointly. Through extensive experiments, the importance of proposed dataset can be verified by the statistics and progressive performances. In addition, our proposed model yields state-of-the-art results, demonstrating its effectiveness.

This paper studies the use of reinforcement learning for conditional text generation, which overcomes the limitation of the prevalent supervised maximum likelihood estimation approach. However, it still suffers from challenges including the large action space and the delayed reward, as the reward can be computed only after an entire sequence is generated. To address these challenges, we propose a method that provides partial rewards for intermediate actions taken on partial sequences. This enables the model to promptly prioritize actions that lead to the generation of more desirable sequences. Our method's key contribution lies in its focus on distinguishing relatively more desirable actions rather than striving to precisely estimate pointwise values for arbitrary partial sequences. Instead, our model learns to discern the relative desirability between pairs of actions, or rank actions in a pairwise manner, only when necessary and feasible. This is materialized in an efficient way by leveraging the prefix tree constructed from the sampled sequences. Experimental results on paraphrase generation and constrained machine translation tasks showcase the effectiveness of our method.

Large, general purpose language models have demonstrated impressive performance across many different conversational domains. While multi-domain language models achieve low overall perplexity, their outputs are not guaranteed to stay within the domain of a given input prompt. This paper proposes \emph{domain privacy} as a novel way to quantify how likely a conditional language model will leak across domains. We also develop policy functions based on token-level domain classification, and propose an efficient fine-tuning method to improve the trained model's domain privacy. Experiments on membership inference attacks show that our proposed method has comparable resiliency to methods adapted from recent literature on differentially private language models.

Target-oriented Multimodal Sentiment Classification (TMSC) aims to incorporate visual modality with text modality to identify the sentiment polarity towards a specific target within a sentence. To address this task, we propose a Visual Elements Mining as Prompts (VEMP) method, which describes the semantic information of visual elements with Text Symbols Embedded in the Image (TSEI), Target-aware Adjective-Noun Pairs (TANPs) and image scene caption, and then transform them into prompts for instruction learning of the model Tk-Instruct. In our VEMP, the text symbols embedded in the image may contain the textual descriptions of fine-grained visual elements, and are extracted as input TSEI; we extract adjective-noun pairs from the image and align them with the target to obtain TANPs, in which the adjectives provide emotional embellishments for the relevant target; finally, to effectively fuse these visual elements with text modality for sentiment prediction, we integrate them to construct instruction prompts for instruction-tuning Tk-Instruct which possesses powerful learning capabilities under instructions. Extensive experimental results show that our method achieves state-of-the-art performance on two benchmark datasets. And further analysis demonstrates the effectiveness of each component of our method.

The abstracts of scientific papers typically contain both premises (e.g., background and observations) and conclusions. Although conclusion sentences are highlighted in structured abstracts, in non-structured abstracts the concluding information is not explicitly marked, which makes the automatic segmentation of conclusions from scientific abstracts a challenging task. In this work, we explore Normalized Mutual Information (NMI) as a means for abstract segmentation. We consider each abstract as a recurrent cycle of sentences and place two segmentation boundaries by greedily optimizing the NMI score between the two segments, assuming that conclusions are strongly semantically linked with preceding premises. On non-structured abstracts, our proposed unsupervised approach GreedyCAS achieves the best performance across all evaluation metrics; on structured abstracts, GreedyCAS outperforms all baseline methods measured by $P_k$. The strong correlation of NMI to our evaluation metrics reveals the effectiveness of NMI for abstract segmentation.

Structured pruning methods have proven effective in reducing the model size and accelerating inference speed in various network architectures such as Transformers. Despite the versatility of encoder-decoder models in numerous NLP tasks, the structured pruning methods on such models are relatively less explored compared to encoder-only models. In this study, we investigate the behavior of the structured pruning of the encoder-decoder models in the decoupled pruning perspective of the encoder and decoder component, respectively. Our findings highlight two insights: (1) the number of decoder layers is the dominant factor of inference speed, and (2) low sparsity in the pruned encoder network enhances generation quality. Motivated by these findings, we propose a simple and effective framework, NASH, that narrows the encoder and shortens the decoder networks of encoder-decoder models. Extensive experiments on diverse generation and inference tasks validate the effectiveness of our method in both speedup and output quality.

Paraphrase Identification (PI), a task of determining whether a pair of sentences express the same meaning, is widely applied in Information Retrieval and Question Answering. Data Augmentation (DA) is proven effective in tackling the PI task. However, the majority of DA methods still suffer from two limitations: inefficiency and poor quality. In this study, we propose the Generative Boosting Training (GBT) approach for PI. GBT designs a boosting learning method for a single model based on the human learning process, utilizing seq2seq model to perform DA on misclassified instances periodically. We conduct experiments on the benchmark corpora QQP and LCQMC, towards both English and Chinese PI tasks. Experimental results show that our method yields significant improvements on a variety of Pre-trained Language Model (PLM) based baselines with good efficiency and effectiveness. It is noteworthy that a single BERT model (with a linear classifier) can outperform the state-of-the-art PI models with the boosting of GBT.

During crisis events, people often use social media platforms such as Twitter to disseminate information about the situation, warnings, advice, and support. Emergency relief organizations leverage such information to acquire timely crisis circumstances and expedite rescue operations. While existing works utilize such information to build models for crisis event analysis, fully-supervised approaches require annotating vast amounts of data and are impractical due to limited response time. On the other hand, semi-supervised models can be biased, performing moderately well for certain classes while performing extremely poorly for others, resulting in substantially negative effects on disaster monitoring and rescue. In this paper, we first study two recent debiasing methods on semi-supervised crisis tweet classification. Then we propose a simple but effective debiasing method, DeCrisisMB, that utilizes a Memory Bank to store and perform equal sampling for generated pseudo-labels from each class at each training iteration. Extensive experiments are conducted to compare different debiasing methods' performance and generalization ability in both in-distribution and out-of-distribution settings. The results demonstrate the superior performance of our proposed method. Our code is available at https://github.com/HenryPengZou/DeCrisisMB.

Engaging with discussion of TV shows online often requires individuals to refrain from consuming show-related content for extended periods to avoid spoilers. While existing research on spoiler detection shows promising results in safeguarding viewers from general spoilers, it fails to address the issue of users abstaining from show-related content during their watch. This is primarily because the definition of a spoiler varies depending on the viewer's progress in the show, and conventional spoiler detection methods lack the granularity to capture this complexity. To tackle this challenge, we propose the task of spoiler matching, which involves assigning an episode number to a spoiler given a specific TV show. We frame this task as semantic text matching and introduce a dataset comprised of comments and episode summaries to evaluate model performance. Given the length of each example, our dataset can also serve as a benchmark for long-range language models.

Recently efforts have been made by social media platforms as well as researchers to detect hateful or toxic language using large language models. However, none of these works aim to use explanation, additional context and victim community information in the detection process. We utilise different prompt variation, input information and evaluate large language models in zero shot setting (without adding any in-context examples). We select two large language models (GPT-3.5 and text-davinci) and three datasets - HateXplain, implicit hate and ToxicSpans. We find that on average including the target information in the pipeline improves the model performance substantially ($\sim20-30\%$) over the baseline across the datasets. There is also a considerable effect of adding the rationales/explanations into the pipeline ($\sim10-20\%$) over the baseline across the datasets. In addition, we further provide a typology of the error cases where these large language models fail to (i) classify and (ii) explain the reason for the decisions they take. Such vulnerable points automatically constitute `jailbreak’ prompts for these models and industry scale safeguard techniques need to be developed to make the models robust against such prompts.

Embodied task completion is a challenge where an agent in a simulated environment must predict environment actions to complete tasks based on natural language instructions and ego-centric visual observations. We propose a variant of this problem where the agent predicts actions at a higher level of abstraction called a plan, which helps make agent actions more interpretable and can be obtained from the appropriate prompting of large language models. We show that multimodal transformer models can outperform language-only models for this problem but fall significantly short of oracle plans. Since collecting human-human dialogues for embodied environments is expensive and time-consuming, we propose a method to synthetically generate such dialogues, which we then use as training data for plan prediction. We demonstrate that multimodal transformer models can attain strong zero-shot performance from our synthetic data, outperforming language-only models trained on human-human data.