Search International and National Patent Collections

1. (WO2018218708) DEEP-LEARNING-BASED PUBLIC OPINION HOTSPOT CATEGORY CLASSIFICATION METHOD

Pub. No.:    WO/2018/218708    International Application No.:    PCT/CN2017/089139
Publication Date: Fri Dec 07 00:59:59 CET 2018 International Filing Date: Wed Jun 21 01:59:59 CEST 2017
IPC: G06F 17/30
Applicants: CHINA UNIVERSITY OF MINING AND TECHNOLOGY
中国矿业大学
Inventors: ZHOU, Yong
周勇
LIU, Bing
刘兵
LIU, Jinxue
刘敬学
WANG, Chongqiu
王重秋
Title: DEEP-LEARNING-BASED PUBLIC OPINION HOTSPOT CATEGORY CLASSIFICATION METHOD
Abstract:
A deep-learning-based public opinion hotspot category classification method, comprising: collecting and pre-processing a training data set, establishing a probability topic presentation model, carrying out two probability distribution presentations, i.e. document-topic and topic-vocabulary, on a text data set, and inputting an obtained topic-vocabulary matrix into a pre-established neural network model for training so as to learn text features; and a network output layer selecting Softmax to carry out normalization processing and classification prediction. The method solves the dimensionality reduction problem of long text public opinion hotspot data, and realises the automatic extraction of deep features of public opinion hotspot information, so that the classification of multiple categories of public opinion hotspots is more accurate.