From Voice to Visualization ― Visual Analysis of Voice Data of Shandong Tax Hotline Based on NLP

Abstract

This paper presents a method through which we can realize the procedure of visual analysis of voice data. We first transform amounts of audio files gained by tax hotline 12366 to text files using Baidu speech recognition service, and divide these text files into words and phrases via ‘ Chinese Segmentation ’. We propose a NLP algorithm that can select keywords from these phrases using ‘ Word2Vec ’ and a contradiction - handling method. These keywords serve as indications of classification model, which is gained from tax service requirement. Then, some of the classification results are visualized in the form of multi - level pie chart, based on which we can find some trends of questions asked by callers. This will raise the service efficien cy and the precision of answer.