Computation and Language (cs.CL)

Abstract: Rhetorical Structure Theory based Discourse Parsing (RST-DP) explores how clauses, sentences, and large text spans compose a whole discourse and presents the rhetorical structure as a hierarchical tree. Existing RST parsing pipelines construct rhetorical structures without the knowledge of document-level content structures, which causes relatively low performance when predicting the discourse relations for large text spans. Recognizing the value of high-level content-related information in facilitating discourse relation recognition, we propose a novel pipeline for RST-DP that incorporates structure-aware news content sentence representations derived from the task of News Discourse Profiling. By incorporating only a few additional layers, this enhanced pipeline exhibits promising performance across various RST parsing metrics.

Abstract: The statistical analysis of large scale legal corpus can provide valuable legal insights. For such analysis one needs to (1) select a subset of the corpus using document retrieval tools, (2) structuralize text using information extraction (IE) systems, and (3) visualize the data for the statistical analysis. Each process demands either specialized tools or programming skills whereas no comprehensive unified "no-code" tools have been available. Especially for IE, if the target information is not predefined in the ontology of the IE system, one needs to build their own system. Here we provide NESTLE, a no code tool for large-scale statistical analysis of legal corpus. With NESTLE, users can search target documents, extract information, and visualize the structured data all via the chat interface with accompanying auxiliary GUI for the fine-level control. NESTLE consists of three main components: a search engine, an end-to-end IE system, and a Large Language Model (LLM) that glues the whole components together and provides the chat interface. Powered by LLM and the end-to-end IE system, NESTLE can extract any type of information that has not been predefined in the IE system opening up the possibility of unlimited customizable statistical analysis of the corpus without writing a single line of code. The use of the custom end-to-end IE system also enables faster and low-cost IE on large scale corpus. We validate our system on 15 Korean precedent IE tasks and 3 legal text classification tasks from LEXGLUE. The comprehensive experiments reveal NESTLE can achieve GPT-4 comparable performance by training the internal IE module with 4 human-labeled, and 192 LLM-labeled examples. The detailed analysis provides the insight on the trade-off between accuracy, time, and cost in building such system.

Abstract: Pre-trained models have achieved success in Chinese Short Text Matching (STM) tasks, but they often rely on superficial clues, leading to a lack of robust predictions. To address this issue, it is crucial to analyze and mitigate the influence of superficial clues on STM models. Our study aims to investigate their over-reliance on the edit distance feature, commonly used to measure the semantic similarity of Chinese text pairs, which can be considered a superficial clue. To mitigate STM models' over-reliance on superficial clues, we propose a novel resampling training strategy called Gradually Learn Samples Containing Superficial Clue (GLS-CSC). Through comprehensive evaluations of In-Domain (I.D.), Robustness (Rob.), and Out-Of-Domain (O.O.D.) test sets, we demonstrate that GLS-CSC outperforms existing methods in terms of enhancing the robustness and generalization of Chinese STM models. Moreover, we conduct a detailed analysis of existing methods and reveal their commonality.

Abstract: Prompt-based classification adapts tasks to a cloze question format utilizing the [MASK] token and the filled tokens are then mapped to labels through pre-defined verbalizers. Recent studies have explored the use of verbalizer embeddings to reduce labor in this process. However, all existing studies require a tuning process for either the pre-trained models or additional trainable embeddings. Meanwhile, the distance between high-dimensional verbalizer embeddings should not be measured by Euclidean distance due to the potential for non-linear manifolds in the representation space. In this study, we propose a tuning-free manifold-based space re-embedding method called Locally Linear Embedding with Intra-class Neighborhood Constraint (LLE-INC) for verbalizer embeddings, which preserves local properties within the same class as guidance for classification. Experimental results indicate that even without tuning any parameters, our LLE-INC is on par with automated verbalizers with parameter tuning. And with the parameter updating, our approach further enhances prompt-based tuning by up to 3.2%. Furthermore, experiments with the LLaMA-7B&13B indicate that LLE-INC is an efficient tuning-free classification approach for the hyper-scale language models.

Abstract: Large Language Models (LLMs) have demonstrated remarkable success in diverse natural language processing (NLP) tasks in general domains. However, LLMs sometimes generate responses with the hallucination about medical facts due to limited domain knowledge. Such shortcomings pose potential risks in the utilization of LLMs within medical contexts. To address this challenge, we propose knowledge-tuning, which leverages structured medical knowledge bases for the LLMs to grasp domain knowledge efficiently and facilitate reliable response generation. We also release cMedKnowQA, a Chinese medical knowledge question-answering dataset constructed from medical knowledge bases to assess the medical knowledge proficiency of LLMs. Experimental results show that the LLMs which are knowledge-tuned with cMedKnowQA, can exhibit higher levels of accuracy in response generation compared with vanilla instruction-tuning and offer a new reliable way for the domain adaptation of LLMs.

Abstract: The application of Large Language Models (LLMs) to the medical domain has stimulated the interest of researchers. Recent studies have focused on constructing Instruction Fine-Tuning (IFT) data through medical knowledge graphs to enrich the interactive medical knowledge of LLMs. However, the medical literature serving as a rich source of medical knowledge remains unexplored. Our work introduces the CALLA dataset to probe LLMs' interactive knowledge acquisition from Chinese medical literature. It assesses the proficiency of LLMs in mastering medical knowledge through a free-dialogue fact-checking task. We identify a phenomenon called the ``fact-following response``, where LLMs tend to affirm facts mentioned in questions and display a reluctance to challenge them. To eliminate the inaccurate evaluation caused by this phenomenon, for the golden fact, we artificially construct test data from two perspectives: one consistent with the fact and one inconsistent with the fact. Drawing from the probing experiment on the CALLA dataset, we conclude that IFT data highly correlated with the medical literature corpus serves as a potent catalyst for LLMs, enabling themselves to skillfully employ the medical knowledge acquired during the pre-training phase within interactive scenarios, enhancing accuracy. Furthermore, we design a framework for automatically constructing IFT data based on medical literature and discuss some real-world applications.

Abstract: As social media becomes increasingly popular, more and more activities related to public health emerge. Current techniques for public health analysis involve popular models such as BERT and large language models (LLMs). However, the costs of training in-domain LLMs for public health are especially expensive. Furthermore, such kinds of in-domain datasets from social media are generally imbalanced. To tackle these challenges, the data imbalance issue can be overcome by data augmentation and balanced training. Moreover, the ability of the LLMs can be effectively utilized by prompting the model properly. In this paper, a novel ALEX framework is proposed to improve the performance of public health analysis on social media by adopting an LLMs explanation mechanism. Results show that our ALEX model got the best performance among all submissions in both Task 2 and Task 4 with a high score in Task 1 in Social Media Mining for Health 2023 (SMM4H)[1]. Our code has been released at https:// github.com/YanJiangJerry/ALEX.

Abstract: Selecting the ``right'' amount of information to include in a summary is a difficult task. A good summary should be detailed and entity-centric without being overly dense and hard to follow. To better understand this tradeoff, we solicit increasingly dense GPT-4 summaries with what we refer to as a ``Chain of Density'' (CoD) prompt. Specifically, GPT-4 generates an initial entity-sparse summary before iteratively incorporating missing salient entities without increasing the length. Summaries generated by CoD are more abstractive, exhibit more fusion, and have less of a lead bias than GPT-4 summaries generated by a vanilla prompt. We conduct a human preference study on 100 CNN DailyMail articles and find that that humans prefer GPT-4 summaries that are more dense than those generated by a vanilla prompt and almost as dense as human written summaries. Qualitative analysis supports the notion that there exists a tradeoff between informativeness and readability. 500 annotated CoD summaries, as well as an extra 5,000 unannotated summaries, are freely available on HuggingFace (https://huggingface.co/datasets/griffin/chain_of_density).

Abstract: Fuzzy Fingerprints have been successfully used as an interpretable text classification technique, but, like most other techniques, have been largely surpassed in performance by Large Pre-trained Language Models, such as BERT or RoBERTa. These models deliver state-of-the-art results in several Natural Language Processing tasks, namely Emotion Recognition in Conversations (ERC), but suffer from the lack of interpretability and explainability. In this paper, we propose to combine the two approaches to perform ERC, as a means to obtain simpler and more interpretable Large Language Models-based classifiers. We propose to feed the utterances and their previous conversational turns to a pre-trained RoBERTa, obtaining contextual embedding utterance representations, that are then supplied to an adapted Fuzzy Fingerprint classification module. We validate our approach on the widely used DailyDialog ERC benchmark dataset, in which we obtain state-of-the-art level results using a much lighter model.

Abstract: Many useful tasks on scientific documents, such as topic classification and citation prediction, involve corpora that span multiple scientific domains. Typically, such tasks are accomplished by representing the text with a vector embedding obtained from a Transformer's single CLS token. In this paper, we argue that using multiple CLS tokens could make a Transformer better specialize to multiple scientific domains. We present Multi2SPE: it encourages each of multiple CLS tokens to learn diverse ways of aggregating token embeddings, then sums them up together to create a single vector representation. We also propose our new multi-domain benchmark, Multi-SciDocs, to test scientific paper vector encoders under multi-domain settings. We show that Multi2SPE reduces error by up to 25 percent in multi-domain citation prediction, while requiring only a negligible amount of computation in addition to one BERT forward pass.

Abstract: Large Language Models (LLMs) have made progress in various real-world tasks, which stimulates requirements for the evaluation of LLMs. Existing LLM evaluation methods are mainly supervised signal-based which depends on static datasets and cannot evaluate the ability of LLMs in dynamic real-world scenarios where deep interaction widely exists. Other LLM evaluation methods are human-based which are costly and time-consuming and are incapable of large-scale evaluation of LLMs. To address the issues above, we propose a novel Deep Interaction-based LLM-evaluation framework. In our proposed framework, LLMs' performances in real-world domains can be evaluated from their deep interaction with other LLMs in elaborately designed evaluation tasks. Furthermore, our proposed framework is a general evaluation method that can be applied to a host of real-world tasks such as machine translation and code generation. We demonstrate the effectiveness of our proposed method through extensive experiments on four elaborately designed evaluation tasks.

Abstract: In recent years, great advances in pre-trained language models (PLMs) have sparked considerable research focus and achieved promising performance on the approach of dense passage retrieval, which aims at retrieving relative passages from massive corpus with given questions. However, most of existing datasets mainly benchmark the models with factoid queries of general commonsense, while specialised fields such as finance and economics remain unexplored due to the deficiency of large-scale and high-quality datasets with expert annotations. In this work, we propose a new task, policy retrieval, by introducing the Chinese Stock Policy Retrieval Dataset (CSPRD), which provides 700+ prospectus passages labeled by experienced experts with relevant articles from 10k+ entries in our collected Chinese policy corpus. Experiments on lexical, embedding and fine-tuned bi-encoder models show the effectiveness of our proposed CSPRD yet also suggests ample potential for improvement. Our best performing baseline achieves 56.1% MRR@10, 28.5% NDCG@10, 37.5% Recall@10 and 80.6% Precision@10 on dev set.

Abstract: Vision-language models (VLMs) have recently demonstrated strong efficacy as visual assistants that can parse natural queries about the visual content and generate human-like outputs. In this work, we explore the ability of these models to demonstrate human-like reasoning based on the perceived information. To address a crucial concern regarding the extent to which their reasoning capabilities are fully consistent and grounded, we also measure the reasoning consistency of these models. We achieve this by proposing a chain-of-thought (CoT) based consistency measure. However, such an evaluation requires a benchmark that encompasses both high-level inference and detailed reasoning chains, which is costly. We tackle this challenge by proposing a LLM-Human-in-the-Loop pipeline, which notably reduces cost while simultaneously ensuring the generation of a high-quality dataset. Based on this pipeline and the existing coarse-grained annotated dataset, we build the CURE benchmark to measure both the zero-shot reasoning performance and consistency of VLMs. We evaluate existing state-of-the-art VLMs, and find that even the best-performing model is unable to demonstrate strong visual reasoning capabilities and consistency, indicating that substantial efforts are required to enable VLMs to perform visual reasoning as systematically and consistently as humans. As an early step, we propose a two-stage training framework aimed at improving both the reasoning performance and consistency of VLMs. The first stage involves employing supervised fine-tuning of VLMs using step-by-step reasoning samples automatically generated by LLMs. In the second stage, we further augment the training process by incorporating feedback provided by LLMs to produce reasoning chains that are highly consistent and grounded. We empirically highlight the effectiveness of our framework in both reasoning performance and consistency.