Document modeling method

Abstract

The invention discloses a document modeling method. A semi-structured document is effectively modeled by designing the method in which words and tag information can be utilized at the same time and the tag type information can be automatically utilized. An autonomous compensation mechanism is utilized, the effects of the tag information of different types on document modeling are learned in a deep Boltzmann machine, and heterogeneous information of different tag types can be taken into full consideration, so that more effective semi-structured document vectors are learned.

Description

technical field [0001] The present invention relates to document processing and modeling technology, and more specifically, to a document modeling processing method. Background technique [0002] The so-called semi-structured documents refer to those document data that contain rich label information, such as web page text with structural information such as category information, title, author, date, etc. With the development of the Internet, more and more semi-structured text data appear in various network applications. Such text data containing tag information is collectively referred to as semi-structured text data (Semi-Structured Documents). How to effectively model this semi-structured text data has become a research hotspot. Traditionally, an effective means of modeling semi-structured document data is to use a topic model (Topicmodel) based on a Bayesian probability graph model. This modeling method is mainly based on the assumption of the bag-of-words model, and s...

AI Technical Summary

Problems solved by technology

Neither topic models based on Bayesian probabilistic graphical models nor deep Boltzmann machines directly model the types of label information

Images