Entity semantic annotation method based on random walk

Abstract

The invention relates to an entity semantic annotation method based on random walk. The method comprises the following steps: firstly, in an offline module, obtaining a steady-state probability matrixof a description text of an entity in a corpus based on a random walk algorithm, namely an entity term correlation score matrix; secondly, in an online query stage, giving the number of selected labeling terms, inputting the number into a user query entity node, finding a row vector corresponding to the query node by utilizing a steady-state probability matrix obtained in an offline module, and sorting all element values in the row vector according to the size to form a label recommendation list corresponding to the query; and finally, selecting the first k terms with relatively large entityterm correlation scores according to user requirements, and recommending the first k terms as feature words of the entity, i.e., semantic tags of the corresponding query entity. According to the method, a random walk model is introduced to solve the problem of data sparsity, semantic annotation can be conducted on an entity on a sparse data set, and compared with an existing keyword annotation method, the recall ratio and the precision ratio are effectively increased.

Description

technical field [0001] The invention relates to a data mining technology, in particular to an entity semantic labeling method based on random walk. Background technique [0002] Tags can effectively organize information on the Internet. Given an entity (document, image, video, etc.), the task of semantic annotation is to recommend several related tag information. There are already many solutions in the existing technology of labeling. Among them, collaborative filtering is the most widely used recommendation technology, but it has problems such as data sparsity and cold start, which directly lead to a great reduction in recommendation quality. In addition, there are keyword tagging methods, which aim to find a few representative words in the original text, but it cannot provide words that may be more representative but do not appear in the document. The tag recommendation of our method differs from keyword extraction in several respects, and the required tag may not appear...

AI Technical Summary

Problems solved by technology

[0004]The present invention is aimed at the sparse real data, the existing method is difficult to label comprehensively, which affects the recall rate and precision ra

Images