A method for generating tweet summaries based on topic relevance

Abstract

The invention discloses a method for generating tweet summaries based on topic correlation, which includes establishing a thesaurus for each topic through the distribution of nouns in each topic; and calculating The correlation between a tweet and a topic; calculate the public recognition degree according to the network interaction information; combine the public recognition degree and the topic correlation to get the final tweet significance; use the maximum marginal correlation algorithm for deduplication processing, Output summary. This method selects tweets as summaries from topic relevance and tweet salience, and controls the redundancy of the final summaries, so that the generated tweet summaries comprehensively consider the summarization topics, diversity, and social identity. As a result, abstracts with higher topic relevance, novelty and summarization are obtained.

Description

Technical field

[0001] The technical field involves text summary technology in natural language processing, which is used to automatically generate a summary of the topic of Twitter speech. Specifically, given a particular topic and several tweet texts, a summary related to that topic is obtained. Background

[0002] With the rapid development of social network media and self-media, abstract research on summarizing massive data has been spawned. Since there is no large-scale public dataset of social network data, most of the current summary studies of social network data are traditional unsupervised methods. Methods based on statistical features, mainly based on the relative position of sentences, word frequency characteristics, etc. to study, such methods are easy to implement, but the obtained features are often relatively simple; based on the method of graph model, such methods regard the sentences in the text as nodes, the similarity score between the texts as the edges betw...

Images