Selecting relevant features for providing improved recommender system through Question Answering task

摘要

In recent years, rapid growth of information available on the web tends user to post massive number of requests across diverse Information Retrieval systems. Question Answering Systems (QAS) play a vital role to provide accurate answers to user questions by relying on several information retrieval components. In this paper, we propose a novel framework for investigating appropriate features to provide recommender system that extract relevant features from user questions. Many linguistic features are generally used for modeling, such as part-of-speech, question focus, location, persona name relative IDF etc. This proposal analyses to extract features from user question sentences, capturing significant query level dependencies between word and phrase in user questions using Natural Language Processing tools. Our system uses questions from TREC data set to measure performance of system. The proposed architecture consists of selecting keyword from the questions and representing using feature sets, calculating query term weights. The study was experimented using TREC data set collections having different features. The experimental results were investigated from twelve features to identify relevant results to questions of different patterns.