Computation and Language (cs.CL)

Abstract: High-quality data is essential for conversational recommendation systems and serves as the cornerstone of the network architecture development and training strategy design. Existing works contribute heavy human efforts to manually labeling or designing and extending recommender dialogue templates. However, they suffer from (i) the limited number of human annotators results in that datasets can hardly capture rich and large-scale cases in the real world, (ii) the limited experience and knowledge of annotators account for the uninformative corpus and inappropriate recommendations. In this paper, we propose a novel automatic dataset synthesis approach that can generate both large-scale and high-quality recommendation dialogues through a data2text generation process, where unstructured recommendation conversations are generated from structured graphs based on user-item information from the real world. In doing so, we comprehensively exploit: (i) rich personalized user profiles from traditional recommendation datasets, (ii) rich external knowledge from knowledge graphs, and (iii) the conversation ability contained in human-to-human conversational recommendation datasets. Extensive experiments validate the benefit brought by the automatically synthesized data under low-resource scenarios and demonstrate the promising potential to facilitate the development of a more effective conversational recommendation system.

Abstract: Given the dynamic nature of toxic language use, automated methods for detecting toxic spans are likely to encounter distributional shift. To explore this phenomenon, we evaluate three approaches for detecting toxic spans under cross-domain conditions: lexicon-based, rationale extraction, and fine-tuned language models. Our findings indicate that a simple method using off-the-shelf lexicons performs best in the cross-domain setup. The cross-domain error analysis suggests that (1) rationale extraction methods are prone to false negatives, while (2) language models, despite performing best for the in-domain case, recall fewer explicitly toxic words than lexicons and are prone to certain types of false positives. Our code is publicly available at: https://github.com/sfschouten/toxic-cross-domain.

Abstract: Relation extraction (RE) aims to extract relations from sentences and documents. Existing relation extraction models typically rely on supervised machine learning. However, recent studies showed that many RE datasets are incompletely annotated. This is known as the false negative problem in which valid relations are falsely annotated as 'no_relation'. Models trained with such data inevitably make similar mistakes during the inference stage. Self-training has been proven effective in alleviating the false negative problem. However, traditional self-training is vulnerable to confirmation bias and exhibits poor performance in minority classes. To overcome this limitation, we proposed a novel class-adaptive re-sampling self-training framework. Specifically, we re-sampled the pseudo-labels for each class by precision and recall scores. Our re-sampling strategy favored the pseudo-labels of classes with high precision and low recall, which improved the overall recall without significantly compromising precision. We conducted experiments on document-level and biomedical relation extraction datasets, and the results showed that our proposed self-training framework consistently outperforms existing competitive methods on the Re-DocRED and ChemDisgene datasets when the training data are incompletely annotated. Our code is released at https://github.com/DAMO-NLP-SG/CAST.

Abstract: Text readability assessment has gained significant attention from researchers in various domains. However, the lack of exploration into corpus compatibility poses a challenge as different research groups utilize different corpora. In this study, we propose a novel evaluation framework, Cross-corpus text Readability Compatibility Assessment (CRCA), to address this issue. The framework encompasses three key components: (1) Corpus: CEFR, CLEC, CLOTH, NES, OSP, and RACE. Linguistic features, GloVe word vector representations, and their fusion features were extracted. (2) Classification models: Machine learning methods (XGBoost, SVM) and deep learning methods (BiLSTM, Attention-BiLSTM) were employed. (3) Compatibility metrics: RJSD, RRNSS, and NDCG metrics. Our findings revealed: (1) Validated corpus compatibility, with OSP standing out as significantly different from other datasets. (2) An adaptation effect among corpora, feature representations, and classification methods. (3) Consistent outcomes across the three metrics, validating the robustness of the compatibility assessment framework. The outcomes of this study offer valuable insights into corpus selection, feature representation, and classification methods, and it can also serve as a beginning effort for cross-corpus transfer learning.

Abstract: Despite the success of ChatGPT, its performances on most NLP tasks are still well below the supervised baselines. In this work, we looked into the causes, and discovered that its subpar performance was caused by the following factors: (1) token limit in the prompt does not allow for the full utilization of the supervised datasets; (2) mismatch between the generation nature of ChatGPT and NLP tasks; (3) intrinsic pitfalls of LLMs models, e.g., hallucination, overly focus on certain keywords, etc. In this work, we propose a collection of general modules to address these issues, in an attempt to push the limits of ChatGPT on NLP tasks. Our proposed modules include (1) a one-input-multiple-prompts strategy that employs multiple prompts for one input to accommodate more demonstrations; (2) using fine-tuned models for better demonstration retrieval; (3) transforming tasks to formats that are more tailored to the generation nature; (4) employing reasoning strategies that are tailored to addressing the task-specific complexity; (5) the self-verification strategy to address the hallucination issue of LLMs; (6) the paraphrase strategy to improve the robustness of model predictions. We conduct experiments on 21 datasets of 10 representative NLP tasks, including question answering, commonsense reasoning, natural language inference, sentiment analysis, named entity recognition, entity-relation extraction, event extraction, dependency parsing, semantic role labeling, and part-of-speech tagging. Using the proposed assemble of techniques, we are able to significantly boost the performance of ChatGPT on the selected NLP tasks, achieving performances comparable to or better than supervised baselines, or even existing SOTA performances.

Abstract: Previous work has predominantly focused on monolingual English semantic parsing. We, instead, explore the feasibility of Chinese semantic parsing in the absence of labeled data for Chinese meaning representations. We describe the pipeline of automatically collecting the linearized Chinese meaning representation data for sequential-to sequential neural networks. We further propose a test suite designed explicitly for Chinese semantic parsing, which provides fine-grained evaluation for parsing performance, where we aim to study Chinese parsing difficulties. Our experimental results show that the difficulty of Chinese semantic parsing is mainly caused by adverbs. Realizing Chinese parsing through machine translation and an English parser yields slightly lower performance than training a model directly on Chinese data.

Abstract: The fast-growing number of research articles makes it problematic for scholars to keep track of the new findings related to their areas of expertise. Furthermore, linking knowledge across disciplines in rapidly developing fields becomes challenging for complex topics like climate change that demand interdisciplinary solutions. At the same time, the rise of Black Box types of text summarization makes it difficult to understand how text relationships are built, let alone relate to existing theories conceptualizing cause-effect relationships and permitting hypothesizing. This work aims to sensibly use Natural Language Processing by extracting variables relations and synthesizing their findings using networks while relating to key concepts dominant in relevant disciplines. As an example, we apply our methodology to the analysis of farmers' adaptation to climate change. For this, we perform a Natural Language Processing analysis of publications returned by Scopus in August 2022. Results show that the use of Natural Language Processing together with networks in a descriptive manner offers a fast and interpretable way to synthesize literature review findings as long as researchers back up results with theory.

Abstract: In efforts to keep up with the rapid progress and use of large language models, gender bias research is becoming more prevalent in NLP. Non-English bias research, however, is still in its infancy with most work focusing on English. In our work, we study how grammatical gender bias relating to politeness levels manifests in Japanese and Korean language models. Linguistic studies in these languages have identified a connection between gender bias and politeness levels, however it is not yet known if language models reproduce these biases. We analyze relative prediction probabilities of the male and female grammatical genders using templates and find that informal polite speech is most indicative of the female grammatical gender, while rude and formal speech is most indicative of the male grammatical gender. Further, we find politeness levels to be an attack vector for allocational gender bias in cyberbullying detection models. Cyberbullies can evade detection through simple techniques abusing politeness levels. We introduce an attack dataset to (i) identify representational gender bias across politeness levels, (ii) demonstrate how gender biases can be abused to bypass cyberbullying detection models and (iii) show that allocational biases can be mitigated via training on our proposed dataset. Through our findings we highlight the importance of bias research moving beyond its current English-centrism.

Abstract: Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) but demand massive GPU resources for training. Lowering the threshold for LLMs training would encourage greater participation from researchers, benefiting both academia and society. While existing approaches have focused on parameter-efficient fine-tuning, which tunes or adds a small number of parameters, few have addressed the challenge of tuning the full parameters of LLMs with limited resources. In this work, we propose a new optimizer, LOw-Memory Optimization (LOMO), which fuses the gradient computation and the parameter update in one step to reduce memory usage. By integrating LOMO with existing memory saving techniques, we reduce memory usage to 10.8% compared to the standard approach (DeepSpeed solution). Consequently, our approach enables the full parameter fine-tuning of a 65B model on a single machine with 8 RTX 3090, each with 24GB memory.

Abstract: Relation Extraction (RE) is a task that identifies relationships between entities in a text, enabling the acquisition of relational facts and bridging the gap between natural language and structured knowledge. However, current RE models often rely on small datasets with low coverage of relation types, particularly when working with languages other than English. In this paper, we address the above issue and provide two new resources that enable the training and evaluation of multilingual RE systems. First, we present SRED$^{\rm FM}$, an automatically annotated dataset covering 18 languages, 400 relation types, 13 entity types, totaling more than 40 million triplet instances. Second, we propose RED$^{\rm FM}$, a smaller, human-revised dataset for seven languages that allows for the evaluation of multilingual RE systems. To demonstrate the utility of these novel datasets, we experiment with the first end-to-end multilingual RE model, mREBEL, that extracts triplets, including entity types, in multiple languages. We release our resources and model checkpoints at https://www.github.com/babelscape/rebel

Abstract: Dialogue systems and large language models (LLMs) have gained considerable attention. However, the direct utilization of LLMs as task-oriented dialogue (TOD) models has been found to underperform compared to smaller task-specific models. Nonetheless, it is crucial to acknowledge the significant potential of LLMs and explore improved approaches for leveraging their impressive abilities. Motivated by the goal of leveraging LLMs, we propose an alternative approach called User-Guided Response Optimization (UGRO) to combine it with a smaller TOD model. This approach uses LLM as annotation-free user simulator to assess dialogue responses, combining them with smaller fine-tuned end-to-end TOD models. By utilizing the satisfaction feedback generated by LLMs, UGRO further optimizes the supervised fine-tuned TOD model. Specifically, the TOD model takes the dialogue history as input and, with the assistance of the user simulator's feedback, generates high-satisfaction responses that meet the user's requirements. Through empirical experiments on two TOD benchmarks, we validate the effectiveness of our method. The results demonstrate that our approach outperforms previous state-of-the-art (SOTA) results.

Abstract: Language models have the potential to assess mental health using social media data. By analyzing online posts and conversations, these models can detect patterns indicating mental health conditions like depression, anxiety, or suicidal thoughts. They examine keywords, language markers, and sentiment to gain insights into an individual's mental well-being. This information is crucial for early detection, intervention, and support, improving mental health care and prevention strategies. However, using language models for mental health assessments from social media has two limitations: (1) They do not compare posts against clinicians' diagnostic processes, and (2) It's challenging to explain language model outputs using concepts that the clinician can understand, i.e., clinician-friendly explanations. In this study, we introduce Process Knowledge-infused Learning (PK-iL), a new learning paradigm that layers clinical process knowledge structures on language model outputs, enabling clinician-friendly explanations of the underlying language model predictions. We rigorously test our methods on existing benchmark datasets, augmented with such clinical process knowledge, and release a new dataset for assessing suicidality. PK-iL performs competitively, achieving a 70% agreement with users, while other XAI methods only achieve 47% agreement (average inter-rater agreement of 0.72). Our evaluations demonstrate that PK-iL effectively explains model predictions to clinicians.

Abstract: In this paper we describe the University of Sheffield's submission to the AmericasNLP 2023 Shared Task on Machine Translation into Indigenous Languages which comprises the translation from Spanish to eleven indigenous languages. Our approach consists of extending, training, and ensembling different variations of NLLB-200. We use data provided by the organizers and data from various other sources such as constitutions, handbooks, news articles, and backtranslations generated from monolingual data. On the dev set, our best submission outperforms the baseline by 11% average chrF across all languages, with substantial improvements particularly for Aymara, Guarani and Quechua. On the test set, we achieve the highest average chrF of all the submissions, we rank first in four of the eleven languages, and at least one of our submissions ranks in the top 3 for all languages.

Abstract: Large Language Models (LLMs) have achieved great success in various natural language tasks. It has aroused much interest in evaluating the specific reasoning capability of LLMs, such as multilingual reasoning and mathematical reasoning. However, as one of the key reasoning perspectives, logical reasoning capability has not yet been thoroughly evaluated. In this work, we aim to bridge those gaps and provide comprehensive evaluations. Firstly, to offer systematic evaluations, this paper selects fifteen typical logical reasoning datasets and organizes them into deductive, inductive, abductive and mixed-form reasoning settings. Considering the comprehensiveness of evaluations, we include three representative LLMs (i.e., text-davinci-003, ChatGPT and BARD) and evaluate them on all selected datasets under zero-shot, one-shot and three-shot settings. Secondly, different from previous evaluations relying only on simple metrics (e.g., accuracy), we propose fine-level evaluations from objective and subjective manners, covering both answers and explanations. Also, to uncover the logical flaws of LLMs, bad cases will be attributed to five error types from two dimensions. Thirdly, to avoid the influences of knowledge bias and purely focus on benchmarking the logical reasoning capability of LLMs, we propose a new dataset with neutral content. It contains 3K samples and covers deductive, inductive and abductive reasoning settings. Based on the in-depth evaluations, this paper finally concludes the ability maps of logical reasoning capability from six dimensions (i.e., correct, rigorous, self-aware, active, oriented and no hallucination). It reflects the pros and cons of LLMs and gives guiding directions for future works.

Abstract: We aimed to investigate the impact of social circumstances on cancer therapy selection using natural language processing to derive insights from social worker documentation. We developed and employed a Bidirectional Encoder Representations from Transformers (BERT) based approach, using a hierarchical multi-step BERT model (BERT-MS) to predict the prescription of targeted cancer therapy to patients based solely on documentation by clinical social workers. Our corpus included free-text clinical social work notes, combined with medication prescription information, for all patients treated for breast cancer. We conducted a feature importance analysis to pinpoint the specific social circumstances that impact cancer therapy selection. Using only social work notes, we consistently predicted the administration of targeted therapies, suggesting systematic differences in treatment selection exist due to non-clinical factors. The UCSF-BERT model, pretrained on clinical text at UCSF, outperformed other publicly available language models with an AUROC of 0.675 and a Macro F1 score of 0.599. The UCSF BERT-MS model, capable of leveraging multiple pieces of notes, surpassed the UCSF-BERT model in both AUROC and Macro-F1. Our feature importance analysis identified several clinically intuitive social determinants of health (SDOH) that potentially contribute to disparities in treatment. Our findings indicate that significant disparities exist among breast cancer patients receiving different types of therapies based on social determinants of health. Social work reports play a crucial role in understanding these disparities in clinical decision-making.

Abstract: Large Language Models (LLMs) have shown remarkable aptitude in code generation but still struggle on challenging programming tasks. Self-repair -- in which the model debugs and fixes mistakes in its own code -- has recently become a popular way to boost performance in these settings. However, only very limited studies on how and when self-repair works effectively exist in the literature, and one might wonder to what extent a model is really capable of providing accurate feedback on why the code is wrong when that code was generated by the same model. In this paper, we analyze GPT-3.5 and GPT-4's ability to perform self-repair on APPS, a challenging dataset consisting of diverse coding challenges. To do so, we first establish a new evaluation strategy dubbed pass@t that measures the pass rate of the tasks against the total number of tokens sampled from the model, enabling a fair comparison to purely sampling-based approaches. With this evaluation strategy, we find that the effectiveness of self-repair is only seen in GPT-4. We also observe that self-repair is bottlenecked by the feedback stage; using GPT-4 to give feedback on the programs generated by GPT-3.5 and using expert human programmers to give feedback on the programs generated by GPT-4, we unlock significant performance gains.

Abstract: Natural Language Inference (NLI) has been a cornerstone task in evaluating language models' inferential reasoning capabilities. However, the standard three-way classification scheme used in NLI has well-known shortcomings in evaluating models' ability to capture the nuances of natural human reasoning. In this paper, we argue that the operationalization of the neutral label in current NLI datasets has low validity, is interpreted inconsistently, and that at least one important sense of neutrality is often ignored. We uncover the detrimental impact of these shortcomings, which in some cases leads to annotation datasets that actually decrease performance on downstream tasks. We compare approaches of handling annotator disagreement and identify flaws in a recent NLI dataset that designs an annotator study based on a problematic operationalization. Our findings highlight the need for a more refined evaluation framework for NLI, and we hope to spark further discussion and action in the NLP community.

Abstract: Large language models have exhibited exceptional performance on various Natural Language Processing (NLP) tasks, leveraging techniques such as the pre-training, and instruction fine-tuning. Despite these advances, their effectiveness in medical applications is limited, due to challenges such as factual inaccuracies, reasoning abilities, and lack grounding in real-world experience. In this study, we present ClinicalGPT, a language model explicitly designed and optimized for clinical scenarios. By incorporating extensive and diverse real-world data, such as medical records, domain-specific knowledge, and multi-round dialogue consultations in the training process, ClinicalGPT is better prepared to handle multiple clinical task. Furthermore, we introduce a comprehensive evaluation framework that includes medical knowledge question-answering, medical exams, patient consultations, and diagnostic analysis of medical records. Our results demonstrate that ClinicalGPT significantly outperforms other models in these tasks, highlighting the effectiveness of our approach in adapting large language models to the critical domain of healthcare.