Bad information identification method and device

Abstract

Embodiments of the invention provide a bad information identification method and a device. The method comprises the steps of obtaining to-be-identified text information; processing the to-be-identified text information by utilizing a syntactic analysis tree to obtain a to-be-identified structure corresponding to the to-be-identified text information and to-be-identified constituent words; judgingwhether the to-be-identified structure is matched with a tree structure of a preset tree rule; if the to-be-identified structure is matched with a tree structure of a preset tree rule, matching the to-be-identified constituent words with constituent words of the preset tree rule; if the to-be-identified constituent words are matched with the constituent words of the preset tree rule, identifying the to-be-identified constituent words; according to the method, sentence analysis is carried out on the text, and matching analysis is carried out on the text, the tree structure of the preset tree rule and the composition words in sequence, so that the problem that the traditional rule-based bad information identification generalization ability is low is solved, and the identification efficiency,accuracy and coverage rate of the bad information can be improved.

Description

technical field [0001] The embodiment of the present invention relates to the technical field of information processing, and in particular to a method and device for identifying bad information. Background technique [0002] With the rapid development of Internet technology, the types of portal websites such as forums and microblogs are increasing day by day, providing convenient channels for information acquisition and speech expression. However, at the same time, many malicious users publish bad information through network channels. The dissemination of bad information will invade the outlook on life, values, and morals of normal users, affect the environment of online communities, harm the interests of others, corrupt the atmosphere of online comments, and hinder normal users from obtaining effective information. In recent years, the state has carried out several professional actions to severely crack down on bad information on the Internet, eradicate the profit chain of...

AI Technical Summary

Problems solved by technology

Among them, the keyword-based method is to judge whether the text is bad information through the matching of the text and the keyword. Its advantage is that the recognition efficiency is high, and its disadvantage is that the accuracy and coverage rate are not high, and the generalization problem cannot be solved; rule-based The method is to construct a rule base by extracting typical representative rules, and judge whether the text is bad information through the matching between the rules and the text. Its advantage is high accuracy, and its disadvantage is that the process of extracting rules will It consumes a lot of human resources and cannot solve the generalization problem; currently, the method based on machine learning is mostly text classification method, which obtains the text representation model through text preprocessing, text feature extraction and feature fusion processing, and then through simple Bayesian, decision tree, random forest and other classification algorithms construct classifiers, and use classifiers to identify bad information. The advantage is that it reduces the consumption of human resources and can solve the problem of generalization to a large extent. However, due to the short and The colloquialism is serious and the number of samples is unbalanced, so the machine learning method cannot achieve good results

Images