An Implicit Emotion Mining Method of User Consultants-oriented
Abstract
Aiming at the characteristics of short texts which are sparse, nonstandard and ambiguous in subject, we present an effective classification method. This paper analyzes this kind of data, and utilize semi-supervised to select significant syntactic features (substring/subsequence), and then uses SVM for text categorization. A machine-learning based emotion analysis method is implemented to mine such implicit emotion. The average accuracy rate and recall rate can reach 84.19%, At last, the method is proven effective by applying in the veritable data of telecommunication field.
Keywords
Text categorization, Semi-supervised, Feature selection
DOI
10.12783/dtssehs/ecemi2020/34672
10.12783/dtssehs/ecemi2020/34672