Knowledge graph-based intention labeling method for mass live broadcast bullet screen data

Abstract

The invention discloses a knowledge graph-based intention labeling method for mass live broadcast bullet screen data. The method comprises the following steps: S1, extracting keywords according to bullet screen information and summarizing dimensions of the keywords; S2, expanding homophones and synonyms of the keywords; S3, combining a plurality of dimensions to serve as a template of a specific intention; S4, carrying out deduplication of bullet screen data and removal of invalid data; S5, extracting viewpoint intentions of the bullet screen data through a template; and S6, carrying out manual checking to remove error data. According to the method, the structure is scientific and reasonable, the use is safe and convenient, the defects that the data volume is large, the labor efficiency is low, the effect of a traditional NLP in bullet screen analysis is poor and the like are overcome, and semantics in the data are defined, so that bullet screen data de-duplication is facilitated, the workload is reduced, and the method is suitable for better popularization and application.

Description

technical field [0001] The invention relates to the technical field of building knowledge thesaurus and data labeling, and specifically relates to a method for labeling intents of a large number of live barrage data based on knowledge graphs. Background technique [0002] Bullet chat can give the audience an illusion of "real-time interaction". Although the sending time of different bullet chat is different, it will only appear at a specific time point in the video, so the bullet chat sent at the same time is basically the same. With the same theme, when participating in the comment, there will be an illusion of commenting with other viewers at the same time, while the traditional player comment system is independent of the player, so the content of the comment is mostly around the entire video, and the topic is not. Strong, and there is no sense of "real-time interaction"; [0003] On the e-commerce service platform, there are customer comments on products or services. The...

AI Technical Summary

Problems solved by technology

[0004] However, these solutions are currently not applicable to colloquial barrage texts. NLP analysis requires complete grammatical sentences, and the main application scenario is written texts. The word vector method has certain effects, but it can only be divided into positive and negative. The keyword method needs to hit the keyword, the loss rate of the spoken language scene is very high, and the results obtained are scattered

Images