Fine-grained emotion analysis improvement method based on emotion word embedding

Abstract

The invention discloses a fine-grained emotion analysis improvement method based on emotion word embedding. The method comprises: obtaining of a text data set, manual labeling and preprocessing; calculating a semantic word vector corresponding to each word in the text data set; obtaining a set of sentiment words; calculating a group of emotion phrases corresponding to each word in the training data set; and calculating an emotion word vector corresponding to each word in the training data set; constructing emotion word embedding corresponding to each word in the training data set; and trainingthe classifier to obtain a fine-grained emotion analysis model. According to the method, priori sentiment knowledge is combined with a word embedding model, and sentiment word embedding suitable forfine-grained sentiment analysis is constructed for sentiment analysis. The emotion information of the word level can be better identified, the emotion of the user can be described more accurately in afine-grained mode, and the method can be used for fine-grained emotion analysis tasks such as consumption habit analysis of the user and comment analysis of the user on commodities.

Description

technical field [0001] The invention belongs to the technical field of Chinese text sentiment analysis, and in particular relates to an improved fine-grained sentiment analysis method based on sentiment word embedding for Chinese short texts such as microblogs and network comments. Background technique [0002] With the rapid development of the Internet, millions of people use social networks every day, such as Weibo, Tieba and other online platforms to express their views on products, services, news, events, etc. Analyzing the opinions or opinions expressed by users is very important for marketing professionals and researchers. Sentiment analysis of Weibo texts has become increasingly important due to the popularity of Weibo in Chinese society. Since 2013, the China Computer Federation (CCF for short) has specially set up the Chinese Weibo sentiment classification evaluation task in the second Natural Language Processing and Chinese Computing Conference (NLPCC for short), ...

AI Technical Summary

Problems solved by technology

However, these methods have certain limitations in sentiment analysis tasks: word embedding implements many NLP tasks by learning low-dimensional continuous value vector representations of words, but the word representation obtained by traditional word embedding methods only contains semantic information in the text corpus , without considering the emotional information in the text corpus

Images