Accuracy evaluation of methods and techniques in Web-based question answering systems: a survey

作者:Asad Ali Shah, Sri Devi Ravana, Suraya Hamid, Maizatul Akmar Ismail

摘要

Question answering (QA) systems answer the queries of users efficiently in the least amount of time. A researcher has to decide which among various methods and techniques available will be used to retrieve accurate answers when developing a QA system. This step creates an overhead before making a selection. The study highlights the methods and techniques that perform well in terms of the accuracy of answers provided. Nine Web-based question answering systems were consulted, and the applied methods and techniques evaluated on the basis of the percentage of questions correctly answered and the mean reciprocal rank evaluation measures. Results were discussed using three key stages involved in a QA system: answer extraction, scoring of answers, and answer aggregation. Results show some techniques have higher accuracy of answers than others. Not all methods in QA systems can improve the accuracy of answers individually, but the methods used in combination obtain greater effect. The results can be used to select methods and techniques optimal for producing highly accurate scores without spending time on benchmarking.

论文关键词:Question answering systems, Answer accuracy, Answer retrieval methods, Information retrieval, Natural language processing

论文评审过程:

论文官网地址:https://doi.org/10.1007/s10115-018-1203-0