Judging Relevance of Candidate Answers in Question Answering System

Building a prototype for answering multiple choice general-interest trivia questions requires an approach for answer selection. In this paper, an approach for answer selection is proposed. The main objectives in this context are: to show that the World Wide Web can be used as data source for answering general-interest questions in a languages such as Arabic and to compare among the performance of different answer selection strategies and their performance with different languages. The different experiments that have been carried out, have shown that the World Wide Web’s redundancy enables the utilization of simple tricks to overcome many troublesome issues in natural language processing, such that for the English language, simple statistical techniques suffice to validate a set of candidate answers while more pre-processing linguistic efforts are still needed for the Arabic language. It has been show that 61% and 56%. Of ''Who wants to be a millionaire'' English version questions and Arabic version questions can be solved by the proposed method respectively.