EMNLP 2024 Papers

This paper proposes a novel approach to evaluate Counter Narrative (CN) generation using a Large Language Model (LLM) as an evaluator. We show that traditional automatic metrics correlate poorly with human judgements and fail to capture the nuanced relationship between generated CNs and human perception. To alleviate this, we introduce a model ranking pipeline based on pairwise comparisons of generated CNs from different models, organized in a tournament-style format. The proposed evaluation method achieves a high correlation with human preference, with a ρ score of 0.88. As an additional contribution, we leverage LLMs as zero-shot CN generators and provide a comparative analysis of chat, instruct, and base models, exploring their respective strengths and limitations. Through meticulous evaluation, including fine-tuning experiments, we elucidate the differences in performance and responsiveness to domain-specific data. We conclude that chat-aligned models in zero-shot are the best option for carrying out the task, provided they do not refuse to generate an answer due to security concerns.

Differential privacy (DP) and federated learning (FL) are combined as advanced privacy-preserving methods when training on-device language models in production mobile keyboard applications. DP-Follow-the-Regularized-Leader (DP-FTRL) algorithms, leveraging correlated noise mechanisms such as tree aggregation or matrix factorization, are widely used in practice for their superior privacy-utility trade-off and compatibility with FL systems. This paper presents a novel variant of DP-FTRL by adapting the recent theoretical advancements of the Buffered Linear Toeplitz (BLT) mechanism to multi-participant scenarios. In the FL setting, our BLT mechanism demonstrates enhanced privacy-utility trade-off and improved memory efficiency than the widely used tree aggregation mechanism. Moreover, BLT achieves comparable privacy and utility to the state-of-the-art banded matrix factorization mechanism, while significantly simplifying usage requirements and reducing memory. The flexibility of the BLT mechanism allows seamless integration with existing DP FL implementations in production environments. We evaluate the BLT-DP-FTRL algorithm on the StackOverflow dataset, serving as a research simulation benchmark, and across four on-device language model tasks in a production FL system. Our empirical results highlight the potential of the BLT mechanism to elevate the practicality and effectiveness of DP in real-world scenarios.

Fine-grained category discovery using only coarse-grained supervision is a cost-effective yet challenging task. Previous training methods focus on aligning query samples with positive samples and distancing them from negatives. They often neglect intra-category and inter-category semantic similarities of fine-grained categories when navigating sample distributions in the embedding space. Furthermore, some evaluation techniques that rely on pre-collected test samples are inadequate for real-time applications. To address these shortcomings, we introduce a method that successfully detects fine-grained clusters of semantically similar texts guided by a novel objective function. The method uses semantic similarities in a logarithmic space to guide sample distributions in the Euclidean space and to form distinct clusters that represent fine-grained categories. We also propose a centroid inference mechanism to support real-time applications. The efficacy of the method is both theoretically justified and empirically confirmed on three benchmark tasks. The proposed objective function is integrated in multiple contrastive learning based neural models. Its results surpass existing state-of-the-art approaches in terms of Accuracy, Adjusted Rand Index and Normalized Mutual Information of the detected fine-grained categories. Code and data are publicly available at https://github.com/changtianluckyforever/F-grained-STAR.

Knowledge graphs (KGs) can provide explainable reasoning for large language models (LLMs), alleviating their hallucination problem. Knowledge graph question answering (KGQA) is a typical benchmark to evaluate the methods enhancing LLMs with KG. Previous methods on KG-enhanced LLM for KGQA either enhance LLMs with KG retrieval in a single round or perform multi-hop KG reasoning in multiple rounds with LLMs. Both of them conduct retrieving and reasoning based solely on the whole original question, without any processing to the question. To tackle this limitation, we propose a framework of KG-enhanced LLM based on question decomposition and atomic retrieval, called KELDaR. We introduce question decomposition tree as the framework for LLM reasoning. This approach extracts the implicit information of reasoning steps within complex questions, serving as a guide to facilitate atomic retrieval on KG targeting the atomic-level simple questions at leaves of the tree. Additionally, we design strategies for atomic retrieval, which extract and retrieve question-relevant KG subgraphs to assist the few-shot LLM in answering atomic-level questions. Experiments on KGQA datasets demonstrate that our framework outperforms existing reasoning-based baselines. And in a low-cost setting without additional training or fine-tuning, our framework achieves competitive or superior results compared to most existing training-based baselines.

We introduce a novel tagging scheme for discontinuous named entity recognition based on an explicit description of the inner structure of discontinuous mentions. We rely on a weighted finite state automaton for both marginal and maximum a posteriori inference. As such, our method is sound in the sense that (1) well-formedness of predicted tag sequences is ensured via the automaton structure and (2) there is an unambiguous mapping between well-formed sequences of tags and (discontinuous) mentions. We evaluate our approach on three English datasets in the biomedical domain, and report comparable results to state-of-the-art while having a way simpler and faster model.

Despite recent advancements in AI and NLP, negotiation remains a difficult domain for AI agents. Traditional game theoretic approaches that have worked well for two-player zero-sum games struggle in the context of negotiation due to their inability to learn human-compatible strategies. On the other hand, approaches that only use human data tend to be domain-specific and lack the theoretical guarantees provided by strategies grounded in game theory. Motivated by the notion of fairness as a criterion for optimality in general sum games, we propose a negotiation framework called FDHC which incorporates fairness into both the reward design and search to learn human-compatible negotiation strategies. Our method includes a novel, RL+search technique called LGM-Zero which leverages a pre-trained language model to retrieve human-compatible offers from large action spaces. Our results show that our method is able to achieve more egalitarian negotiation outcomes and improve negotiation quality.

In this paper, we present a study on the impact of so-called multiword expressions (MWEs) and multiword named entities (NEs) on the performance of Chinese-English machine translation (MT) systems. Built on an extended version of the data from the WMT22 Metrics Shared Task (with extra labels of 9 types of Chinese MWEs, and 19 types of Chinese multiword NEs) which includes scores and error annotations provided by human experts, we make further extraction of MWE- and NE-related translation errors. By investigating the human evaluation scores and the error rates on each category of MWEs and NEs, we find that: 1) MT systems tend to perform significantly worse on Chinese sentences with most kinds of MWEs and NEs; 2) MWEs and NEs which make up of about twenty percent of tokens, i.e. characters in Chinese, result in one-third of translation errors; 3) for 13 categories of MWEs and NEs, the error rates exceed 50% with the highest to be 84.8%. Based on the results, we emphasize that MWEs and NEs are still a bottleneck issue for MT and special attention to MWEs and NEs should be paid to further improving the performance of MT systems.

Length-control summarization aims to condense long texts into a short one within a certain length limit. Previous approaches often use autoregressive (AR) models and treat the length requirement as a soft constraint, which may not always be satisfied. In this study, we propose a novel length-control decoding algorithm based on the directed acyclic Transformer (DAT). Our approach allows for multiple plausible sequence fragments and predicts a path to connect them. In addition, we propose a Sequence Maximum a Posteriori (Seq-MAP) decoding algorithm that marginalizes different possible paths and finds the most probable summary satisfying the length budget. Our algorithm is based on beam search, which further facilitates a reranker for performance improvement. Experimental results on the Gigaword dataset demonstrate our state-of-the-art performance for length-control summarization.

Effective summarisation evaluation metrics enable researchers and practitioners to compare different summarisation systems efficiently. Estimating the effectiveness of an automatic evaluation metric, termed meta-evaluation, is a critically important research question. In this position paper, we review recent meta-evaluation practices for summarisation evaluation metrics and find that (1) evaluation metrics are primarily meta-evaluated on datasets consisting of examples from news summarisation datasets, and (2) there has been a noticeable shift in research focus towards evaluating the faithfulness of generated summaries. We argue that the time is ripe to build more diverse benchmarks that enable the development of more robust evaluation metrics and analyze the generalization ability of existing evaluation metrics. In addition, we call for research focusing on user-centric quality dimensions that consider the generated summary’s communicative goal and the role of summarisation in the workflow.

Extracting product information is crucial for informed business decisions and strategic planning across multiple industries. However, recent methods relying only on large language models (LLMs) are resource-intensive and computationally prohibitive due to website structure differences and numerous non-product pages. To address these challenges, we propose a novel modular method that leverages low-cost classification models to filter out company web pages, significantly reducing computational costs. Our approach consists of three modules: web page crawling, product page classification using efficient machine learning models, and product information extraction using LLMs on classified product pages. We evaluate our method on a new dataset of about 7000 product and non-product web pages, achieving a 6-point improvement in F1-score, 95% reduction in computational time, and 87.5% reduction in cost compared to end-to-end LLMs. Our research demonstrates the effectiveness of our proposed low-cost classification module to identify web pages containing product information, making product information extraction more effective and cost-efficient.

In many scientific fields, large language models (LLMs) have revolutionized the way text and other modalities of data (e.g., molecules and proteins) are handled, achieving superior performance in various applications and augmenting the scientific discovery process. Nevertheless, previous surveys on scientific LLMs often concentrate on one or two fields or a single modality. In this paper, we aim to provide a more holistic view of the research landscape by unveiling cross-field and cross-modal connections between scientific LLMs regarding their architectures and pre-training techniques. To this end, we comprehensively survey over 260 scientific LLMs, discuss their commonalities and differences, as well as summarize pre-training datasets and evaluation tasks for each field and modality. Moreover, we investigate how LLMs have been deployed to benefit scientific discovery. Resources related to this survey are available at https://github.com/yuzhimanhua/Awesome-Scientific-Language-Models.

The rapid advancement of foundation models (FMs) across language, image, audio, and video domains has shown remarkable capabilities in diverse tasks. However, the proliferation of FMs brings forth a critical challenge: the potential to generate hallucinated outputs, particularly in high-stakes applications. The tendency of foundation models to produce hallucinated content arguably represents the biggest hindrance to their widespread adoption in real-world scenarios, especially in domains where reliability and accuracy are paramount. This survey paper presents a comprehensive overview of recent developments that aim to identify and mitigate the problem of hallucination in FMs, spanning text, image, video, and audio modalities. By synthesizing recent advancements in detecting and mitigating hallucination across various modalities, the paper aims to provide valuable insights for researchers, developers, and practitioners. Essentially, it establishes a clear framework encompassing definition, taxonomy, and detection strategies for addressing hallucination in multimodal foundation models, laying the foundation for future research and development in this pivotal area.

Pretrained language models (PLMs) display impressive performances and have captured the attention of the NLP community.Establishing best practices in pretraining has, therefore, become a major focus of NLP research, especially since insights gained from monolingual English models may not necessarily apply to more complex multilingual models.One significant caveat of the current state of the art is that different works are rarely comparable: they often discuss different parameter counts, training data, and evaluation methodology.This paper proposes a comparison of multilingual pretraining objectives in a controlled methodological environment. We ensure that training data and model architectures are comparable, and discuss the downstream performances across 6 languages that we observe in probing and fine-tuning scenarios.We make two key observations: (1) the architecture dictates which pretraining objective is optimal; (2) multilingual translation is a very effective pretraining objective under the right conditions.We make our code, data, and model weights available at https://github.com/Helsinki-NLP/lm-vs-mt.

Few-shot intent detection is a challenging task, particularly in scenarios involving multiple labels and diverse domains. This paper presents a novel prototype learning approach that combines the label synset augmentation and the coarse-to-fine prototype distillation for multi-label few-shot intent detection. To tackle the data scarcity issue and the lack of information for unseen domains, we propose to enhance the representations of utterances with label synset augmentation and refine the prototypes by distilling the coarse domain knowledge from a universal teacher model. To solve the multilingual intent detection in real-world dialogue systems, we fine-tune a cross-lingual teacher model to make our method fast adapt to different languages and re-annotate two non-English task-oriented dialogue datasets CrossWOZ and JMultiWOZ in multi-label form. Experimental results on one English and two non-English datasets demonstrate that our approach significantly outperforms existing methods in terms of accuracy and generalization across different domains.

Toxic online political discourse has become prevalent, where scholars debate about its impact to Democratic processes. This work presents a large-scale study of political incivility on Twitter. In line with theories of political communication, we differentiate between harsh ‘impolite’ style and intolerant substance. We present a dataset of 13K political tweets in the U.S. context, which we collected and labeled by those categories using crowd sourcing. Our dataset and results shed light on hostile political discourse focused on partisan conflicts in the U.S. The evaluation of state-of-the-art classifiers illustrates the challenges involved in political incivility detection, which often requires high-level semantic and social understanding. Nevertheless, performing incivility detection at scale, we are able to characterise its distribution across individual users and geopolitical regions, where our findings align and extend existing theories of political communication. In particular, we find that roughly 80% of the uncivil tweets are authored by 20% of the users, where users who are politically engaged are more inclined to use uncivil language. We further find that political incivility exhibits network homophily, and that incivility is more prominent in highly competitive geopolitical regions. Our results apply to both uncivil style and substance.

Authorship attribution aims to identify the origin or author of a document. Traditional approaches have heavily relied on manual features and fail to capture long-range correlations, limiting their effectiveness. Recent advancements leverage text embeddings from pre-trained language models, which require significant fine-tuning on labeled data, posing challenges in data dependency and limited interpretability. Large Language Models (LLMs), with their deep reasoning capabilities and ability to maintain long-range textual associations, offer a promising alternative. This study explores the potential of pre-trained LLMs in one-shot authorship attribution, specifically utilizing Bayesian approaches and probability outputs of LLMs. Our methodology calculates the probability that a text entails previous writings of an author, reflecting a more nuanced understanding of authorship. By utilizing only pre-trained models such as Llama-3-70B, our results on the IMDb and blog datasets show an impressive 85% accuracy in one-shot authorship classification across ten authors. Our findings set new baselines for one-shot authorship analysis using LLMs and expand the application scope of these models in forensic linguistics. This work also includes extensive ablation studies to validate our approach.

Large Language Models (LLMs) have garnered significant attention due to their remarkable ability to process information across various languages. Despite their capabilities, they exhibit inconsistencies in handling identical queries in different languages, presenting challenges for further advancement. This paper introduces a method to enhance the multilingual performance of LLMs by aggregating knowledge from diverse languages. This approach incorporates a low-resource knowledge detector specific to a language, a strategic language selection process, and mechanisms for answer replacement and integration. Our extensive experiments demonstrate notable performance improvements, particularly in reducing the performance disparity across languages. An ablation study confirms that each component of our method significantly contributes to these enhancements. This research highlights the inherent potential of LLMs to harmonize multilingual capabilities and offers valuable insights for further exploration.

Understanding satire and humor is a challenging task for even current Vision-Language models. In this paper, we propose the challenging tasks of Satirical Image Detection (detecting whether an image is satirical), Understanding (generating the reason behind the image being satirical), and Completion (given one half of the image, selecting the other half from 2 given options, such that the complete image is satirical) and release a high-quality dataset ***YesBut***, consisting of 2547 images, 1084 satirical and 1463 non-satirical, containing different artistic styles, to evaluate those tasks. Each satirical image in the dataset depicts a normal scenario, along with a conflicting scenario which is funny or ironic. Despite the success of current Vision-Language Models on multimodal tasks such as Visual QA and Image Captioning, our benchmarking experiments show that such models perform poorly on the proposed tasks on the ***YesBut*** Dataset in Zero-Shot Settings w.r.t both automated as well as human evaluation. Additionally, we release a dataset of 119 real, satirical photographs for further research.