Entity relationship classification method based on probability distribution self-adaptation

Abstract

The invention discloses an entity relationship classification method based on probability distribution self-adaptation, which comprises the following steps: firstly, acquiring source domain data and target domain data, and training word vector characteristics of the two domain data and position characteristics of entity words; then, through forward propagation calculation of a deep neural networkmodel, acquiring advanced feature representation of the data in the two fields; calculating an edge probability distribution difference and a conditional probability distribution difference between domain data advanced feature representations; then, calculating A-distance between the two probability distributions, endowing the two probability distributions with different learning weights, and automatically updating the weights in the training process; and calculating through a loss function, updating network parameters through back propagation in combination with the loss of the probability distribution difference, and finally training to obtain a classification model for classification testing. The domain adaptation is achieved by reducing the distribution difference between data in different domains, and the target domain classification effect is improved by utilizing a large amount of data in a source domain.

Description

technical field [0001] The invention relates to the technical field of machine learning and natural language processing, in particular to an entity relationship classification method based on probability distribution self-adaptation. Background technique [0002] In recent years, with the vigorous development of Internet technology, especially the popularity of mobile network devices, the number of Internet users worldwide has exceeded 4 billion, and the amount of data generated has shown geometric growth. Information in the network has different organizational forms, and unstructured or semi-structured text is the most common information carrier. Information extraction technology refers to the extraction of information that meets specific needs from text data with different levels of structure, and plays an important role in the processing of massive text information and the construction of knowledge bases. [0003] The entity relationship extraction task is to extract the...

AI Technical Summary

Problems solved by technology

[0007] In view of this, the present invention provides an adaptive entity relationship classification method based on probability distribution to solve or at least partly solve the technical problem of poor classification effect existing in the methods in the prior art

Images