Short Text Classification Method Combining Word Vector and WTTM

Abstract

Aimed at the problem that the characteristics of news headlines are sparse and the semantic relationship between words and words is weak, which led to the difficulty of obtaining good results in traditional text classification methods, a short text based on word vector and WTTM model is proposed. A classification method that models the topic of short text from the co-occurrence of words. Firstly, the Word2Vec tool is used to train the word vector in the short text corpus and the average word vector is synthesized by the additive averaging method. Then the theme of the short text corpus is modeled by the WTTM model to obtain the topic extended feature vector. Finally, the average word vector and the topic extended feature vector are merged. And used a random forest model to construct a classifier for classification. Compared with other short text classification methods, the short text classification method based on word vector and WTTM model has improved accuracy, recall rate and F1 value, which verified the feasibility of the method.