A lexical vector-based emotion analysis method for automotive product reviews

Abstract

The invention discloses a lexical vector-based emotion analysis method for automotive product reviews. The method comprises the following steps: acquiring automobile domain keyword database and preprocessing the original text to form text to be labeled; obtaining the automobile domain keyword database and preprocessing the automobile domain keyword database to form text to be labeled. For the sample to be marked, the sample is selected according to the time span standard and the vehicle type grade standard. Word2vec model is used to obtain the word vector of the text to be analyzed, and the high-dimensional sentence vector is obtained from the mean value of the word vector. PCA principal component analysis is used to reduce the dimension of high-dimensional sentence vector, and the reduceddimension sentence vector and the result of affective classification are used as feature training SVM classifier. The classifier is used to analyze the emotion of the new text, and the result of emotion analysis is generated. The invention is based on the automobile product knowledge ontology, pertinently builds the domain thesaurus, effectively labels the comment text in the domain, and obtainsmore accurate emotion analysis results by using the word vectorization and the machine learning model.

Description

technical field [0001] The invention relates to the fields of text mining and natural language processing, in particular to a sentiment analysis method for car product reviews based on word vectors. Background technique [0002] In recent years, the automobile industry in China is developing rapidly in a gratifying manner. According to statistics from the Ministry of Public Security, by the end of 2017, the number of motor vehicles in the country reached 310 million, including 217 million cars; the number of motor vehicle drivers reached 385 million, including 442 million car drivers [1] . Continuous income growth and optimal control of the cost of automotive products have brought more and more families into access to automotive products. In 2017, 33.52 million motor vehicles were newly registered with the public security traffic management department, of which 28.13 million were newly registered vehicles, both hitting record highs [2] . Automobile products have graduall...

AI Technical Summary

Problems solved by technology

[0008] 1) The breadth and depth of the dictionary in the professional field of automobiles is not enough, and it cannot cover all the structures and functions of automobile products;

[0009] 2) There is a lack of dictionary support for user-generated content. When users express their subjective feelings on the Internet, they will use abbreviations of automobile products (such as "first four and rear four"), nicknames (such as "Ma Liu", "four sons") ”), etc., the lack of a dictionary that considers the Internet context will reduce the accuracy of sentiment analysis results;

[0011] 4) The machine learning model input characterized by TF-IDF (term frequency-inverse text frequency index) cannot well reflect the semantics and context of the text

Images