User opinion extraction method and system

Abstract

The invention discloses a user opinion extraction method and system. The method comprises the steps that text data is processed according to a word vector model, and a word vector set of a corpus in the text data is obtained; a sentence is processed according to word vectors to obtain sentence vectors; clustering analysis is performed on the sentence vectors to obtain an opinion cluster of a user;a central user comment in the opinion cluster is extracted to obtain a user opinion; and a user emotion is given through a depth model composed of a convolutional neural network and a recurrent neural network. Through the user opinion extraction method and system, a typical opinion fed back by the user on a certain product or service and emotion analysis of the opinion can be automatically extracted, therefore, the heavy task of manual summarization extraction which is mostly adopted currently is avoided, and production efficiency is substantially improved.

Description

Technical field [0001] The present invention relates to the field of big data, in particular to a method and system for extracting user opinions. Background technique [0002] With the rapid development of the Internet, there are more and more websites in segmented fields, such as websites for services such as automobiles, catering, and accommodation. Many users in the subdivision field, along with their own practical experience, will express their own views on their services. User feedback is of great significance to the product or service provider, especially to track the user’s public opinion trend on the product or service and subsequent improvements to the product or service. [0003] Opinion extraction of user feedback data is the focus and difficulty of big data analysis. Through automatic extraction methods, most users’ concerns are extracted from massive user feedback data to further improve products or services. [0004] In the prior art, technical solutions for extractin...

AI Technical Summary

Problems solved by technology

[0007] Manually collect feedback from users of a product, and use some auxiliary tools, such as excel, to manually summarize it. Extracting user opinions requires a lot of human resources, and when the amount of data is relatively large, such as Due to the limitations of hardware resources such as memory and computing power, self-media with a huge number of users, such as Official Accounts and Toutiao Accounts, makes it difficult to fully consider all sample data and make a summary

[0008] It is difficult to apply the artificial summary method to automatic realization. For example, the popular public opinion monitoring nowadays requires the system to grasp the trend of public opinion in a certain product or field in real time. By automatically capturing all relevant data on the Internet and performing real-time analysis and processing, the obtained The overall public opinion of the masses on the product, and the artificial summary method obviously cannot meet the application scenario

[0009] The existing method of obtaining typical opinions of users through BOW model combined with clustering algorithm is more suitable for long texts, such as news data, so that the text features will not be too sparse, but user comment data or forum post data are all short sentences Indicates that the feature space mapped to tens of thousands of dimensions is too sparse to be effectively processed by subsequent models, so the effect of this scheme is average

Images