EMNLP 2023 论文列表

Large language models (LLMs) have recently reached an impressive level of linguistic capability, prompting comparisons with human language skills. However, there have been relatively few systematic inquiries into the linguistic capabilities of the latest generation of LLMs, and those studies that do exist (i) ignore the remarkable ability of humans to generalize, (ii) focus only on English, and (iii) investigate syntax or semantics and overlook other capabilities that lie at the heart of human language, like morphology. Here, we close these gaps by conducting the first rigorous analysis of the morphological capabilities of ChatGPT in four typologically varied languages (specifically, English, German, Tamil, and Turkish). We apply a version of Berko's (1958) wug test to ChatGPT, using novel, uncontaminated datasets for the four examined languages. We find that ChatGPT massively underperforms purpose-built systems, particularly in English. Overall, our results---through the lens of morphology---cast a new light on the linguistic capabilities of ChatGPT, suggesting that claims of human-like language skills are premature and misleading.

Large language models (LLMs) have showcased their capability with few-shot inference known as in-context learning. However, in-domain demonstrations are not always readily available in real scenarios, leading to cross-domain in-context learning. Besides, LLMs are still facing challenges in long-tail knowledge in unseen and unfamiliar domains. The above limitations demonstrate the necessity of Unsupervised Domain Adaptation (UDA). In this paper, we study the UDA problem under an in-context learning setting to adapt language models from the source domain to the target domain without any target labels. The core idea is to retrieve a subset of cross-domain elements that are the most similar to the query, and elicit language model to adapt in an in-context manner by learning both target domain distribution and the discriminative task signal simultaneously with the augmented cross-domain in-context examples. We devise different prompting and training strategies, accounting for different LM architectures to learn the target distribution via language modeling. With extensive experiments on Sentiment Analysis (SA) and Named Entity Recognition (NER) tasks, we thoroughly study the effectiveness of ICL for domain transfer and demonstrate significant improvements over baseline models.

With the rapid development of neural network applications in NLP, model robustness problem is gaining more attention. Different from computer vision, the discrete nature of texts makes it more challenging to explore robustness in NLP. Therefore, in this paper, we aim to connect discrete perturbations with continuous perturbations, therefore we can use such connections as a bridge to help understand discrete perturbations in NLP models. Specifically, we first explore how to connect and measure the correlation between discrete perturbations and continuous perturbations. Then we design a regression task as a PerturbScore to learn the correlation automatically. Through experimental results, we find that we can build a connection between discrete and continuous perturbations and use the proposed PerturbScore to learn such correlation, surpassing previous methods used in discrete perturbation measuring. Further, the proposed PerturbScore can be well generalized to different datasets, perturbation methods, indicating that we can use it as a powerful tool to study model robustness in NLP.

Compared to traditional single-turn ad-hoc retrieval, conversational retrieval needs to handle the multi-turn conversation and understand the user’s real query intent. However, most existing methods simply fine-tune the pre-trained ad-hoc retriever on limited supervised data, making it challenging for the retriever to fully grasp the entirety of the conversation. In this paper, we find that large language models (LLMs) can accurately discover the user’s query intent from the complex conversation context and provide the supervised signal to instruct the retriever in an unsupervised manner. Therefore, we propose a novel method termed InstructoR to Instruct unsupervised conversational dense Retrieval with LLMs. We design an unsupervised training framework that employs LLMs to estimate the session-passage relevance score as the soft label to guide the retriever's training. Specially, we devise three instructing strategies from context, query and response perspectives to calculate the relevance score more precisely, including conversational retrieval as conversation generation, question rewrite as latent variable and question response as posterior guide. Experimental results show InstructoR can bring significant improvements across various ad-hoc retrievers, even surpassing the current supervised state-of-the-art method. We also demonstrate the effectiveness of our method under low-resource and zero-shot settings. Our code is publicly available at https://github.com/jinzhuoran/InstructoR/.

Human evaluation in often considered to be the gold standard method of evaluating a Natural Language Generation system. However, whilst its importance is accepted by the community at large, the quality of its execution is often brought into question. In this position paper, we argue that the generation of more esoteric forms of language - humour, irony and sarcasm - constitutes a subdomain where the characteristics of selected evaluator panels are of utmost importance, and every effort should be made to report demographic characteristics wherever possible, in the interest of transparency and replicability. We support these claims with an overview of each language form and an analysis of examples in terms of how their interpretation is affected by different participant variables. We additionally perform a critical survey of recent works in NLG to assess how well evaluation procedures are reported in this subdomain, and note a severe lack of open reporting of evaluator demographic information, and a significant reliance on crowdsourcing platforms for recruitment.

As language models are applied to an increasing number of real-world applications, understanding their inner workings has become an important issue in model trust, interpretability, and transparency. In this work we show that representation dissimilarity measures, which are functions that measure the extent to which two model's internal representations differ, can be a valuable tool for gaining insight into the mechanics of language models. Among our insights are: (i) an apparent asymmetry in the internal representations of model using SoLU and GeLU activation functions, (ii) evidence that dissimilarity measures can identify and locate generalization properties of models that are invisible via in-distribution test set performance, and (iii) new evaluations of how language model features vary as width and depth are increased. Our results suggest that dissimilarity measures are a promising set of tools for shedding light on the inner workings of language models.

A salient characteristic of pre-trained language models (PTLMs) is a remarkable improvement in their generalization capability and emergence of new capabilities with increasing model capacity and pre-training dataset size. Consequently, we are witnessing the development of enormous models pushing the state-of-the-art. It is, however, imperative to realize that this inevitably leads to prohibitively long training times, extortionate computing costs, and a detrimental environmental impact. Significant efforts are underway to make PTLM training more efficient through innovations in model architectures, training pipelines, and loss function design, with scant attention being paid to optimizing the utility of training data. The key question that we ask is whether it is possible to train PTLMs by employing only highly informative subsets of the training data while maintaining downstream performance? Building upon the recent progress in informative data subset selection, we show how we can employ submodular optimization to select highly representative subsets of the training corpora and demonstrate that the proposed framework can be applied to efficiently train multiple PTLMs (BERT, BioBERT, GPT-2) using only a fraction of data. Further, we perform a rigorous empirical evaluation to show that the resulting models achieve up to $\sim99\%$ of the performance of the fully-trained models. We made our framework publicly available at \url{https://github.com/Efficient-AI/ingenious}.

Transformer-based models have achieved state-of-the-art performance on numerous NLP applications. However, long documents which are prevalent in real-world scenarios cannot be efficiently processed by transformers with the vanilla self-attention module due to their quadratic computation complexity and limited length extrapolation ability. Instead of tackling the computation difficulty for self-attention with sparse or hierarchical structures, in this paper, we investigate the use of State-Space Models (SSMs) for long document classification tasks. We conducted extensive experiments on six long document classification datasets, including binary, multi-class, and multi-label classification, comparing SSMs (with and without pre-training) to self-attention-based models. We also introduce the SSM-pooler model and demonstrate that it achieves comparable performance while being on average 36\% more efficient. Additionally our method exhibits higher robustness to the input noise even in the extreme scenario of 40\%.

Efficient knowledge retrieval plays a pivotal role in ensuring the success of end-to-end task-oriented dialogue systems by facilitating the selection of relevant information necessary to fulfill user requests. However, current approaches generally integrate knowledge retrieval and response generation, which poses scalability challenges when dealing with extensive knowledge bases. Taking inspiration from open-domain question answering, we propose a retriever-generator architecture that harnesses a retriever to retrieve pertinent knowledge and a generator to generate system responses. Due to the lack of retriever training labels, we propose relying on feedback from the generator as pseudo-labels to train the retriever. To achieve this, we introduce a dual-feedback mechanism that generates both positive and negative feedback based on the output of the generator. Our method demonstrates superior performance in task-oriented dialogue tasks, as evidenced by experimental results on three benchmark datasets.

Metaphor identification aims at understanding whether a given expression is used figuratively in context. However, in this paper we show how existing metaphor identification datasets can be gamed by fully ignoring the potential metaphorical expression or the context in which it occurs. We test this hypothesis in a variety of datasets and settings, and show that metaphor identification systems based on language models without complete information can be competitive with those using the full context. This is due to the construction procedures to build such datasets, which introduce unwanted biases for positive and negative classes. Finally, we test the same hypothesis on datasets that are carefully sampled from natural corpora and where this bias is not present, making these datasets more challenging and reliable.

Language Models (LMs) have shown impressive performance in various natural language tasks. However, when it comes to natural language reasoning, LMs still face challenges such as hallucination, generating incorrect intermediate reasoning steps, and making mathematical errors. Recent research has focused on enhancing LMs through *self-improvement* using feedback. Nevertheless, existing approaches relying on a single generic feedback source fail to address the diverse error types found in LM-generated reasoning chains. In this work, we propose **Multi-Aspect Feedback**, an iterative refinement framework that integrates multiple feedback modules, including frozen LMs and external tools, each focusing on a specific error category. Our experimental results demonstrate the efficacy of our approach to addressing several errors in the LM-generated reasoning chain and thus improving the overall performance of an LM in several reasoning tasks. We see an improvement of up to 20% in Mathematical Reasoning and up to 18% in Logical Entailment.

The automatic Brain CT reports generation can improve the efficiency and accuracy of diagnosing cranial diseases. However, current methods are limited by 1) coarse-grained supervision: the training data in image-text format lacks detailed supervision for recognizing subtle abnormalities, and 2) coupled cross-modal alignment: visual-textual alignment may be inevitably coupled in a coarse-grained manner, resulting in tangled feature representation for report generation. In this paper, we propose a novel Pathological Graph-driven Cross-modal Alignment (PGCA) model for accurate and robust Brain CT report generation. Our approach effectively decouples the cross-modal alignment by constructing a Pathological Graph to learn fine-grained visual cues and align them with textual words. This graph comprises heterogeneous nodes representing essential pathological attributes (i.e., tissue and lesion) connected by intra- and inter-attribute edges with prior domain knowledge. Through carefully designed graph embedding and updating modules, our model refines the visual features of subtle tissues and lesions and aligns them with textual words using contrastive learning. Extensive experimental results confirm the viability of our method. We believe that our PGCA model holds the potential to greatly enhance the automatic generation of Brain CT reports and ultimately contribute to improved cranial disease diagnosis.

Behavioral testing offers a crucial means of diagnosing linguistic errors and assessing capabilities of NLP models. However, applying behavioral testing to machine translation (MT) systems is challenging as it generally requires human efforts to craft references for evaluating the translation quality of such systems on newly generated test cases. Existing works in behavioral testing of MT systems circumvent this by evaluating translation quality without references, but this restricts diagnosis to specific types of errors, such as incorrect translation of single numeric or currency words. In order to diagnose general errors, this paper proposes a new Bilingual Translation Pair Generation based Behavior Testing (BTPGBT) framework for conducting behavioral testing of MT systems. The core idea of BTPGBT is to employ a novel bilingual translation pair generation (BTPG) approach that automates the construction of high-quality test cases and their pseudoreferences. Experimental results on various MT systems demonstrate that BTPGBT could provide comprehensive and accurate behavioral testing results for general error diagnosis, which further leads to several insightful findings. Our code and data are available at https: //github.com/wujunjie1998/BTPGBT.

Open-domain fact verification is the task of verifying claims in natural language texts against extracted evidence. FEVEROUS is a benchmark that requires extracting and integrating both unstructured and structured evidence to verify a given claim. Previous models suffer from low recall of structured evidence extraction, i.e., table extraction and cell selection. In this paper, we propose a simple but effective method to enhance the extraction of structured evidence by leveraging the row and column semantics of tables. Our method comprises two components: (i) a coarse-grained table extraction module that selects tables based on rows and columns relevant to the claim and (ii) a fine-grained cell selection graph that combines both formats of evidence and enables multi-hop and numerical reasoning. We evaluate our method on FEVEROUS and achieve an evidence recall of $60.01\%$ on the test set, which is $6.14\%$ higher than the previous state-of-the-art performance. Our results demonstrate that our method can extract tables and select cells effectively, and provide better evidence sets for verdict prediction. Our code is released at https://github.com/WilliamZR/see-st

Retrieval-augmented methods are successful in the standard scenario where the retrieval space is sufficient; whereas in the few-shot scenario with limited retrieval space, this paper shows it is non-trivial to put them into practice. First, it is impossible to retrieve semantically similar examples by using an off-the-shelf metric and it is crucial to learn a task-specific retrieval metric; Second, our preliminary experiments demonstrate that it is difficult to optimize a plausible metric by minimizing the standard cross-entropy loss. The in-depth analyses quantitatively show minimizing cross-entropy loss suffers from the weak supervision signals and the severe gradient vanishing issue during the optimization. To address these issues, we introduce two novel training objectives, namely EM-L and R-L, which provide more task-specific guidance to the retrieval metric by the EM algorithm and a ranking-based loss, respectively. Extensive experiments on $10$ datasets prove the superiority of the proposed retrieval augmented methods on the performance.

Real-world Temporal Knowledge Graphs keep growing with time and new entities and facts emerge continually, necessitating a model that can extrapolate to future timestamps and transfer knowledge for new components. Therefore, our work first dives into this more realistic issue, lifelong TKG reasoning, where existing methods can only address part of the challenges. Specifically, we formulate lifelong TKG reasoning as a temporal-path-based reinforcement learning (RL) framework. Then, we add temporal displacement into the action space of RL to extrapolate for the future and further propose a temporal-rule-based reward shaping to guide the training. To transfer and update knowledge, we design a new edge-aware message passing module, where the embeddings of new entities and edges are inductive. We conduct extensive experiments on three newly constructed benchmarks for lifelong TKG reasoning. Experimental results show the outperforming effectiveness of our model against all well-adapted baselines.

Keyphrase Generation (KPG) is a longstanding task in NLP with widespread applications. The advent of sequence-to-sequence (seq2seq) pre-trained language models (PLMs) has ushered in a transformative era for KPG, yielding promising performance improvements. However, many design decisions remain unexplored and are often made arbitrarily. This paper undertakes a systematic analysis of the influence of model selection and decoding strategies on PLM-based KPG. We begin by elucidating why seq2seq PLMs are apt for KPG, anchored by an attention-driven hypothesis. We then establish that conventional wisdom for selecting seq2seq PLMs lacks depth: (1) merely increasing model size or performing task-specific adaptation is not parameter-efficient; (2) although combining in-domain pre-training with task adaptation benefits KPG, it does partially hinder generalization. Regarding decoding, we demonstrate that while greedy search achieves strong F1 scores, it lags in recall compared with sampling-based methods. Based on these insights, we propose DeSel, a likelihood-based decode-select algorithm for seq2seq PLMs. DeSel improves greedy search by an average of 4.7% semantic F1 across five datasets. Our collective findings pave the way for deeper future investigations into PLM-based KPG.

Existing evaluations of entity linking systems often say little about how the system is going to perform for a particular application. There are two fundamental reasons for this. One is that many evaluations only use aggregate measures (like precision, recall, and F1 score), without a detailed error analysis or a closer look at the results. The other is that all of the widely used benchmarks have strong biases and artifacts, in particular: a strong focus on named entities, an unclear or missing specification of what else counts as an entity mention, poor handling of ambiguities, and an over- or underrepresentation of certain kinds of entities. We provide a more meaningful and fair in-depth evaluation of a variety of existing end-to-end entity linkers. We characterize their strengths and weaknesses and also report on reproducibility aspects. The detailed results of our evaluation can be inspected under https://elevant.cs.uni-freiburg.de/emnlp2023. Our evaluation is based on several widely used benchmarks, which exhibit the problems mentioned above to various degrees, as well as on two new benchmarks, which address the problems mentioned above. The new benchmarks can be found under https://github.com/ad-freiburg/fair-entity-linking-benchmarks.

This paper introduces the \textbf{C}hinese \textbf{E}ssay \textbf{D}iscourse \textbf{C}oherence \textbf{C}orpus (\textbf{CEDCC}), a multi-task dataset for assessing discourse coherence. Existing research tends to focus on isolated dimensions of discourse coherence, a gap which the CEDCC addresses by integrating coherence grading, topical continuity, and discourse relations. This approach, alongside detailed annotations, captures the subtleties of real-world texts and stimulates progress in Chinese discourse coherence analysis. Our contributions include the development of the CEDCC, the establishment of baselines for further research, and the demonstration of the impact of coherence on discourse relation recognition and automated essay scoring. The dataset and related codes is available at \url{https://github.com/cubenlp/CEDCC_corpus}.