Text writing real-time recommendation method based on topic model and user behavior data

Abstract

The invention discloses a topic model and user behavior data-based text writing real-time recommendation method, which comprises the following steps of: carrying out LDA topic model modeling on a historical document to obtain topic characteristics of the historical document; obtaining text data of a current writing document, and obtaining topK keywords and word frequency information of the current document by calculating a TextRank algorithm; obtaining a union set from the document keywords and the topK keywords of all the topics; calculating the cosine similarity of the document vector after union and the subject term vector of each subject to generate a subject vector of the document, calculating the cosine similarity of the subject vector of the document and the subject vector of each historical document, and sorting to obtain a similarity matrix and normalizing the similarity matrix; acquiring user behavior data written by historical documents and modeling, calculating to obtain personalized scores of the historical documents and normalizing the personalized scores, comprehensively scoring all the documents in combination with topic similarity, obtaining a recommendation list which is sequenced according to the comprehensive scores and sequencing in a progressive decrease manner, and recommending TOP-N documents.

Description

technical field [0001] The present invention relates to the field of information technology, specifically to the field of data mining and text recommendation, and more specifically to a real-time text writing recommendation method based on topic models and user behavior data. Background technique [0002] When writing official documents and other texts, the writer will produce a large number of historical writing texts during the writing process, and these historical texts will help the creator's writing in terms of writing ideas and writing content. However, how to find data that matches the current writing text from the massive historical texts and make recommendations is a problem that needs to be solved. [0003] The current mainstream text recommendation solution in the industry is to use the LDA topic model for model training and prediction to calculate the similarity between the current writing text and the historical document vector, and then use the similarity TOP-N...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to solve the problems of poor short text recommendation effect, long model prediction time and no personalized recommendation for different writers under the above-mentioned prior art, and proposes a real-time text writing method based on topic models and user behavior data. recommended method

Images