Regarding the Semantics Combined with N-gram between Question and Answer Statements to Obtain Factor Score

For the purpose of resolving the issues of mismatches between the intentions of questioners and respondents at Question and Answer (Q&A) sites, nine factors of impressions for Q&A statements were experimentally obtained through factor analysis. Factor scores were then estimated by using the feature values of statements through multiple regression analysis. The factor scores estimated were subsequently used for finding respondents who would be expected to give appropriate answers to a posted question. Then, in order to consider the semantics of the Q&A statement, word2vecs were appended to the feature values of statements to estimate factor scores. It was demonstrated that word2vec showed the potential of improving estimation accuracy, regardless of the feature values based on 2-gram or syntactic information. For the method using N -gram, it would be necessary to analyze whether longer units than 2-gram could be applicable to this methodology. Thus, this paper aims to append N-gram beyond 3-gram to word2vec and closing sentence expressions as the feature values of statements. The analysis result, taking cross-validation into consideration, has shown that word2vec, as well as units of either 3-gram, 4-gram, or 5-gram, would be applicable to estimating factor scores of Q&A statements.