Research on Chinese-English Hybrid Rhetorical Question Recognition Model and Corpus Construction of Intelligent Web Text

Abstract

With the popularity of English in China, Chinese-English mixed rhetorical question has become a common expression in China. Mixed Chinese-English rhetorical questions have rich emotional overtones, and if they can be correctly identified, they improve the results of sentiment analysis and other tasks. Using semi-supervised learning and active learning methods, a semi-automatic collection of the rhetorical corpus is proposed to construct a Chinese-English rhetorical corpus of web text. Based on the corpus, the characteristics of Chinese-English mixed rhetorical questions are analyzed from the aspects of semantic features, positional features, and syntactic path features to carry out a rhetorical question recognition experiment. Experimental results show that the rhetorical question corpus constructed from online texts trains a rhetorical question recognition model with high performance, and the accuracy, recall, and F1 values of the model are higher than 90%. At the same time, the experimental results verify the effectiveness of syntactic path features and location features in identifying rhetorical questions.