Information mining method and apparatus

Abstract

Embodiments of the present invention provide an information mining method and apparatus. The method comprises the following steps of: mining each query statement of each specific category from a search log; giving the particular class of seed entities; generating an expression template corresponding to each query statement of the specific category according to the seed entity of the specific category and each query statement; according to various query sentences and corresponding expression templates, extracting high-frequency query sentences and high-frequency expression templates from the search log; By using the search log of the user as the data source, the obtained high-frequency expression of high-frequency sentences are rich and can cover all kinds of expression habits of users, which can include the content that cannot be covered by artificial enriched templates such as colloquial expression.

Description

technical field [0001] The invention relates to the technical field of information retrieval, in particular to an information mining method and device. Background technique [0002] In the human-computer interaction system, users express various needs for robot interaction. Existing template-based parsing modules need to have a full amount of user query query statements (query), in order to improve the recall rate of user understanding and parsing accuracy. These user expressions have the following characteristics, which cause many problems in the use of traditional artificial enrichment rules and vocabulary. [0003] (1) There are various ways of expression. Users of the same problem have various forms of expression, and different users have various expression habits. In this case, artificial enrichment construction cannot cover all expressions. [0004] (2) The expression is colloquial, and the user's expression form is very colloquial, which cannot be covered by artific...

AI Technical Summary

Problems solved by technology

[0003] (1) There are various ways of expression. Users of the same problem have various forms of expression, and different users have various expression habits. In this case, artificial enrichment construction cannot cover all expressions

[0004] (2) The expression is colloquial, and the user's expression form is very colloquial, and artificially enriched templates cannot be covered

[0005] (3) The number of vocabularies in each dimension is huge, and it is impossible to construct such a huge vocabulary manually

[0006] Due to the above characteristics of user expression, if artificial enrichment rules and vocabulary are used, there will be problems such as high time and labor costs, low efficiency, and poor analysis effect, which will lead to poor user understanding module and poor human-computer interaction experience

In addition, the enrichment vocabulary cannot enrich a large-scale full-scale vocabulary, resulting in a low parsing recall rate

Enrichment expression methods cannot enrich large-scale full expression templates and colloquial expressions, resulting in low analysis recall and accuracy, inability to understand user expressions, and inability to provide accurate answers, resulting in low user satisfaction

Images