[4G3-GS-2] Machine learning: General

[4G3-GS-2-03] Estimation of Factor Scores of Q&A Statements through Syntactic Analysis

〇Yuya Yokoyama¹ (1. Advanced Institute of Industrial Technology)

[[Online]]

Keywords:Multiple regression analysis, Morphological analysis, N-gram, Factor score

Among the diverse web services along with the widespread of the Internet, there is an increasing number of Question and Answer (Q&A) site users. Q&A site is a form of community where users mutually post questions and answers, and is utilized as a database accumulating vast knowledge as well as a place of resolving various troubles and consultations. However, with the increase of users, major social issues are occurring, e.g. sustaining losses due to not obtaining appropriate answers. Therefore, this work aims to resolve the mismatch between questioner and a respondent, and develop a method to introduce users who can appropriately answer questions. So far nine factors have experimentally been obtained through impression evaluation experiment. In this paper, through multiple regression analysis, the factor scores of any Q&A statements are estimated; factor scores of the 60 Q&A statements utilized for the experiment are set as respondent variables, while feature values extracted through two kinds of syntactic analyses (morphological analysis and N-gram) are set as explanatory variables. As a result, good estimation results are obtained for both of morphological analysis and N-gram.