To explain the predicted answers and evaluate the reasoning abilities of models, several studies have utilized underlying reasoning (UR) tasks in multi-hop question answering (QA) datasets. However, it remains an open question as to how effective UR tasks are for the QA task when training models on both tasks in an end-to-end manner. In this study, we address this question by analyzing the effectiveness of UR tasks (including both sentence-level and entity-level tasks) in three aspects: (1) QA performance, (2) reasoning shortcuts, and (3) robustness. While the previous models have not been explicitly trained on an entity-level reasoning prediction task, we build a multi-task model that performs three tasks together: sentence-level supporting facts prediction, entity-level reasoning prediction, and answer prediction. Experimental results on 2WikiMultiHopQA and HotpotQA-small datasets reveal that (1) UR tasks can improve QA performance. Using four debiased datasets that are newly created, we demonstrate that (2) UR tasks are helpful in preventing reasoning shortcuts in the multi-hop QA task. However, we find that (3) UR tasks do not contribute to improving the robustness of the model on adversarial questions, such as sub-questions and inverted questions. We encourage future studies to investigate the effectiveness of entity-level reasoning in the form of natural language questions (e.g., sub-question forms).

Medical experts may use Artificial Intelligence (AI) systems with greater trust if these are supported by contextual explanations that let the practitioner connect system inferences to their context of use. However, their importance in improving model usage and understanding has not been extensively studied. Hence, we consider a comorbidity risk prediction scenario and focus on contexts regarding the patients clinical state, AI predictions about their risk of complications, and algorithmic explanations supporting the predictions. We explore how relevant information for such dimensions can be extracted from Medical guidelines to answer typical questions from clinical practitioners. We identify this as a question answering (QA) task and employ several state-of-the-art LLMs to present contexts around risk prediction model inferences and evaluate their acceptability. Finally, we study the benefits of contextual explanations by building an end-to-end AI pipeline including data cohorting, AI risk modeling, post-hoc model explanations, and prototyped a visual dashboard to present the combined insights from different context dimensions and data sources, while predicting and identifying the drivers of risk of Chronic Kidney Disease - a common type-2 diabetes comorbidity. All of these steps were performed in engagement with medical experts, including a final evaluation of the dashboard results by an expert medical panel. We show that LLMs, in particular BERT and SciBERT, can be readily deployed to extract some relevant explanations to support clinical usage. To understand the value-add of the contextual explanations, the expert panel evaluated these regarding actionable insights in the relevant clinical setting. Overall, our paper is one of the first end-to-end analyses identifying the feasibility and benefits of contextual explanations in a real-world clinical use case.

Despite recent progress in abstractive summarization, models often generate summaries with factual errors. Numerous approaches to detect these errors have been proposed, the most popular of which are question answering (QA)-based factuality metrics. These have been shown to work well at predicting summary-level factuality and have potential to localize errors within summaries, but this latter capability has not been systematically evaluated in past research. In this paper, we conduct the first such analysis and find that, contrary to our expectations, QA-based frameworks fail to correctly identify error spans in generated summaries and are outperformed by trivial exact match baselines. Our analysis reveals a major reason for such poor localization: questions generated by the QG module often inherit errors from non-factual summaries which are then propagated further into downstream modules. Moreover, even human-in-the-loop question generation cannot easily offset these problems. Our experiments conclusively show that there exist fundamental issues with localization using the QA framework which cannot be fixed solely by stronger QA and QG models.

This paper presents ViDeBERTa, a new pre-trained monolingual language model for Vietnamese, with three versions - ViDeBERTa_xsmall, ViDeBERTa_base, and ViDeBERTa_large, which are pre-trained on a large-scale corpus of high-quality and diverse Vietnamese texts using DeBERTa architecture. Although many successful pre-trained language models based on Transformer have been widely proposed for the English language, there are still few pre-trained models for Vietnamese, a low-resource language, that perform good results on downstream tasks, especially Question answering. We fine-tune and evaluate our model on three important natural language downstream tasks, Part-of-speech tagging, Named-entity recognition, and Question answering. The empirical results demonstrate that ViDeBERTa with far fewer parameters surpasses the previous state-of-the-art models on multiple Vietnamese-specific natural language understanding tasks. Notably, ViDeBERTa_base with 86M parameters, which is only about 23% of PhoBERT_large with 370M parameters, still performs the same or better results than the previous state-of-the-art model. Our ViDeBERTa models are available at: https://github.com/HySonLab/ViDeBERTa.

Controversy is widespread online. Previous studies mainly define controversy based on vague assumptions of its relation to sentiment such as hate speech and offensive words. This paper introduces the first question-answering dataset that defines content controversy by user perception, i.e., votes from plenty of users. It contains nearly 10K questions, and each question has a best answer and a most controversial answer. Experimental results reveal that controversy detection in question answering is essential and challenging, and there is no strong correlation between controversy and sentiment tasks.

Conversational Question Answering (ConvQA) models aim at answering a question with its relevant paragraph and previous question-answer pairs that occurred during conversation multiple times. To apply such models to a real-world scenario, some existing work uses predicted answers, instead of unavailable ground-truth answers, as the conversation history for inference. However, since these models usually predict wrong answers, using all the predictions without filtering significantly hampers the model performance. To address this problem, we propose to filter out inaccurate answers in the conversation history based on their estimated confidences and uncertainties from the ConvQA model, without making any architectural changes. Moreover, to make the confidence and uncertainty values more reliable, we propose to further calibrate them, thereby smoothing the model predictions. We validate our models, Answer Selection-based realistic Conversation Question Answering, on two standard ConvQA datasets, and the results show that our models significantly outperform relevant baselines. Code is available at: https://github.com/starsuzi/AS-ConvQA.

A long-standing challenge for search and conversational assistants is query intention detection in ambiguous queries. Asking clarifying questions in conversational search has been widely studied and considered an effective solution to resolve query ambiguity. Existing work have explored various approaches for clarifying question ranking and generation. However, due to the lack of real conversational search data, they have to use artificial datasets for training, which limits their generalizability to real-world search scenarios. As a result, the industry has shown reluctance to implement them in reality, further suspending the availability of real conversational search interaction data. The above dilemma can be formulated as a cold start problem of clarifying question generation and conversational search in general. Furthermore, even if we do have large-scale conversational logs, it is not realistic to gather training data that can comprehensively cover all possible queries and topics in open-domain search scenarios. The risk of fitting bias when training a clarifying question retrieval/generation model on incomprehensive dataset is thus another important challenge. In this work, we innovatively explore generating clarifying questions in a zero-shot setting to overcome the cold start problem and we propose a constrained clarifying question generation system which uses both question templates and query facets to guide the effective and precise question generation. The experiment results show that our method outperforms existing state-of-the-art zero-shot baselines by a large margin. Human annotations to our model outputs also indicate our method generates 25.2\% more natural questions, 18.1\% more useful questions, 6.1\% less unnatural and 4\% less useless questions.

Question Generation (QG) is a fundamental NLP task for many downstream applications. Recent studies on open-book QG, where supportive answer-context pairs are provided to models, have achieved promising progress. However, generating natural questions under a more practical closed-book setting that lacks these supporting documents still remains a challenge. In this work, we propose a new QG model for this closed-book setting that is designed to better understand the semantics of long-form abstractive answers and store more information in its parameters through contrastive learning and an answer reconstruction module. Through experiments, we validate the proposed QG model on both public datasets and a new WikiCQA dataset. Empirical results show that the proposed QG model outperforms baselines in both automatic evaluation and human evaluation. In addition, we show how to leverage the proposed model to improve existing question-answering systems. These results further indicate the effectiveness of our QG model for enhancing closed-book question-answering tasks.

Background: Extractive question-answering (EQA) is a useful natural language processing (NLP) application for answering patient-specific questions by locating answers in their clinical notes. Realistic clinical EQA can have multiple answers to a single question and multiple focus points in one question, which are lacking in the existing datasets for development of artificial intelligence solutions. Objective: Create a dataset for developing and evaluating clinical EQA systems that can handle natural multi-answer and multi-focus questions. Methods: We leveraged the annotated relations from the 2018 National NLP Clinical Challenges (n2c2) corpus to generate an EQA dataset. Specifically, the 1-to-N, M-to-1, and M-to-N drug-reason relations were included to form the multi-answer and multi-focus QA entries, which represent more complex and natural challenges in addition to the basic one-drug-one-reason cases. A baseline solution was developed and tested on the dataset. Results: The derived RxWhyQA dataset contains 96,939 QA entries. Among the answerable questions, 25% require multiple answers, and 2% ask about multiple drugs within one question. There are frequent cues observed around the answers in the text, and 90% of the drug and reason terms occur within the same or an adjacent sentence. The baseline EQA solution achieved a best f1-measure of 0.72 on the entire dataset, and on specific subsets, it was: 0.93 on the unanswerable questions, 0.48 on single-drug questions versus 0.60 on multi-drug questions, 0.54 on the single-answer questions versus 0.43 on multi-answer questions. Discussion: The RxWhyQA dataset can be used to train and evaluate systems that need to handle multi-answer and multi-focus questions. Specifically, multi-answer EQA appears to be challenging and therefore warrants more investment in research.

A deployed question answering (QA) model can easily fail when the test data has a distribution shift compared to the training data. Robustness tuning (RT) methods have been widely studied to enhance model robustness against distribution shifts before model deployment. However, can we improve a model after deployment? To answer this question, we evaluate test-time adaptation (TTA) to improve a model after deployment. We first introduce COLDQA, a unified evaluation benchmark for robust QA against text corruption and changes in language and domain. We then evaluate previous TTA methods on COLDQA and compare them to RT methods. We also propose a novel TTA method called online imitation learning (OIL). Through extensive experiments, we find that TTA is comparable to RT methods, and applying TTA after RT can significantly boost the performance on COLDQA. Our proposed OIL improves TTA to be more robust to variation in hyper-parameters and test distributions over time.