Semantic analysis method and device

Abstract

The invention discloses a semantic analysis method and device. The semantic analysis method includes the steps that S1, input voice data are acquired, and text transformation is performed on the voice data to generate text data; S2, entity identification is performed on the text data to acquire an entity type tag in the text data; S3, the text data are transformed into text data with an entity type tag according to entity information in the text data and the corresponding entity type tag; S4, the text data with the entity type tag are matched with a database to generate a semantic analysis result. By means of embodiment of the semantic analysis method and device, influences on semantic analysis from noise can be lowered, the entity information in the text data and the entity type tag corresponding to the entity information can be accurately acquired, and therefore analysis accuracy is improved.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a semantic analysis method and device. Background technique [0002] Spoken language semantic analysis is an essential part of voice interaction. It mainly analyzes the intention of the user input, extracts the knowledge that can express the intention, and converts it into a structured data format that the machine can understand. [0003] At present, the method of character string matching between preset rule strings and speech-converted text strings is mainly used to analyze the semantics. Specifically, keyword combinations can be compiled into regular expressions, the extracted function parameters can be represented as wildcards, and a mapping relationship can be established between the regular expressions and corresponding services, thereby realizing semantic analysis. For example, the keyword combination is "(to) (.+) (send / return) (information / short message)", and...

AI Technical Summary

Problems solved by technology

[0004] However, using wildcards to extract function parameters is easy to misunderstand noise, for example: "Send a message to my dear Andy Lau", the correct function parameter should be "Andy Lau", and the extracted function parameter is "My dear Andy Lau", It will affect the subsequent processing, and then affect the overall analysis effect

In addition, when the input text contains multiple function parameters, it is impossible to parse out the meaning of each parameter, such as "query the weather of a certain day and place", the text will be converted into "(query)(.+)(of )? (.+)(weather)", so it is difficult to distinguish which parameter represents the time and which parameter represents the place name, resulting in a low parsing success rate

Images