In a question answering (QA) system, each user interaction with the system is different and since there are a variety of arguably correct answers to complex questions, identifying factors for improving the quality of the retrieved answer is difficult. This research aims to develop a framework that identifies predictive variables for the best quality answer in a QA system. It was found that accuracy, completeness and relevance were predictors of best answer. We believe that these findings can serve to guide future developments in the answer extraction modules in the QA systems.
Purpose -The purpose of this study is to examine the predictors of high-quality answers in a community-driven question answering service (Yahoo! Answers). Design/methodology/approach -The identified predictors were organised into two categories: social and content features. Social features refer to the community aspects of the users and are extracted from explicit user interaction and feedback. Content features refer to the intrinsic and extrinsic content quality of answers that could be used to select the high-quality answers. In total the framework built in this study comprises 17 features from two categories. Based on a randomly selected dataset of 1,600 question-answer pairs from Yahoo! Answers, high-quality answer predictors were identified. Findings -The results of the analysis showed the importance of content appraisal features over social and textual content features. The features identified as strongly associated with high-quality answers include positive votes, completeness, presentation, reliability and accuracy. Features weakly associated with high-quality answers were high frequency words, answer length, and best answers answered. Features related to the asker's user history were found not to be associated with high-quality answers. Practical implications -This work could help in the reuse of answers for new questions. The study identified features that most influence the selection of high-quality answers. Hence they could be used to select high-quality answers for answering similar questions posed by users in the future. When a new question is posed, similar questions are first identified, and the answers for these questions are extracted and routed to the proposed quality framework for identifying high-quality answers. Based on the overall quality index computed, the high-quality answer could be returned to the asker. Originality/value -Previous studies in identifying high-quality answers were conducted using either of two approaches. First using social and textual content features found in community-driven question answering services and second using content appraisal features by thorough assessment of answer quality provided by experts. However no study had integrated both approaches. Hence this study addresses this gap by developing an integrated generalisable framework to identify features that influence high-quality answers.
Background The World Health Organization notes that diabetes, a chronic disease, is a silent epidemic, and by 2020 there will be a 54% rise in the total number of individuals diagnosed with this disease. These are alarming figures that have significant repercussions for the quality of life of individuals and their families as well as for the financial stress of health care systems globally. Early detection and proactive management of diabetes is essential. The Diamond solution provides diabetes self-management by enabling patients to send details about their blood sugar readings at specific times to their nominated care coordinator to receive recommendations for diet and exercise and insulin titration. Objective The aim of the study was to assess the usability, acceptability, and fidelity of the Diamond diabetes monitoring device for patients with gestational diabetes mellitus (GDM). Specifically assessed were (1) patient compliance, (2) patient satisfaction, (3) level of glycemic control achieved, and (4) health professional satisfaction. Methods Using a design science research perspective, the Diamond diabetes monitoring device solution was adapted to the Australian health care environment. Once the solution was deemed fit for purpose by the director of the OB/GYN clinical institute and on securing all relevant ethics approvals, a 2-period 2-arm nonblinded crossover clinical trial was conducted for 8 weeks total time with crossover at 4 weeks to establish proof of concept, usability, and fidelity. The patient perspective was assessed by using structured questionnaires at 4 specific stages of the project, while the clinician perspective was captured via semistructured interviews and unstructured questionnaires. Results The 10 patients studied reported preferring standard care with the technology solution to standard care alone. Further, all clinicians involved concurred that the technology solution greatly assisted their ability to provide higher value patient-centered care. They also noted that it was extremely helpful for assisting in systematically monitoring glucose levels and any/all changes and trends. Conclusions Based on these initial findings, we offer a holistic pervasive approach to enable the achievement of value-based, patient-centered care in chronic disease management. Key lessons include the importance when designing such solutions to focus on the two primary user groups (patients and clinicians).
Community-driven question-answering (CQA) services on the Internet let users share content in the form of questions and answers. Usually, questions attract multiple answers of varying quality from other users. A new approach aims to identify high-quality answers from candidate answers to questions that are semantically similar to the new question. Toward that end, the authors developed and tested a quality framework comprising social, textual, and content-appraisal features of usergenerated answers in CQA services. Logistic-regression analysis revealed that content-appraisal features were the strongest predictor of quality. These features include dimensions such as comprehensiveness, truthfulness, and practicality.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.