Application of 5-gram to Obtain Factor Scores of Japanese Q&A Statements

In order to clarify the issue of mismatches between the questioners and respondents of Question and Answer (Q&A) sites, nine factors of impressions for Q&A statements were obtained through impression evaluation experiments. Then through multiple regression analysis, factor scores were estimated with the usage of feature values of statements. The factor scores estimated and obtained were subsequently utilized for detecting respondents who would be expected to appropriately answer a posted question. Nevertheless, this method has considerably been dependent on the syntactic information extracted through morphological analysis. Moreover, this method has a vital downside of requiring numerous explanatory variables and complicated multiple regression equation to estimate factor scores. Therefore, the principle has been shifted to applying N-gram over morphological analysis. In this paper, 5-gram is applied to the feature values, in a similar way as the previous analysis using shorter N-grams. Further analysis has shown that 5-gram would be also applicable to the method. In terms of estimation accuracy, N-grams also outscore morphological analysis; above all 2-gram and 3-gram show the best accuracy. Hence, it could be suggested that N-gram should play more important role in estimating factor scores than mere morphological analysis.