Mining question-answer pairs from web forum: a survey of challenges and resolutions

Abstract

Internet forums, which are also known as discussion boards, are popular web applications. Members of the board discuss issues and share ideas to form a community within the board, and as a result generate huge amount of content on different topics on daily basis. Interest in information extraction and knowledge discovery from such sources has been on the increase in the research community. A number of factors are limiting the potentiality of mining knowledge from forums. Lexical chasm or lexical gap that renders some Natural Language Processing techniques (NLP) less effective, Informal tone that creates noisy data, drifting of discussion topic that prevents focused mining and asynchronous issue that makes it difficult to establish post-reply relationship are some of the problems that need to be addressed. This survey introduces these challenges within the framework of question answering. The survey provides description of the problems; cites and explores useful publications to the reader for further examination; provides an overview of resolution strategies and findings relevant to the challenges.