Speech recognition test data generation method based on context in user comments

Abstract

The invention discloses a speech recognition test data generation method based on context in user comments. The method comprises the following steps of: extracting context factors of speech recognition and priority and semantic association relationships among the factors from the user comments of a plurality of speech recognition APPs, and extracting value domains of the context factors accordingto a selected data set to construct a context classification tree model. The model can modify original test data in the value domains according to specific test requirements to generate more availabletest data. Through adoption of the speech recognition test data generation method, the problem of severe insufficiency of the test data due to high updating frequency of an intelligent system is solved. Meanwhile, the value domains of the context factors of a real data set are utilized, and a method for rapidly generating real test data is provided for a test speech recognition system, so that the speech recognition test efficiency is increased effectively. Meanwhile, a developer can understand the performance of the speech recognition system more easily, so that the system can be upgraded and updated more specifically.

Description

technical field [0001] The invention belongs to the technical field of intelligent software testing, and in particular relates to a method for generating test data for speech recognition based on context in user comments. Background technique [0002] With the development of artificial intelligence and big data technology, more and more products based on artificial intelligence have emerged in the market, such as face recognition, speech recognition, machine translation and so on. On the basis of traditional software, intelligent software integrates intelligent functional attributes, which undoubtedly brings many new problems and difficulties to the testing of intelligent software, and also puts forward greater market and research needs for intelligent software testing. The current rapid iteration of intelligent software generally has problems such as insufficient test data and insufficient reliability, and it is difficult to meet the test requirements. Meanwhile, the test ...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the deficiencies of the above-mentioned prior art, the object of the present invention is to provide a method for generating speech recognition test data based on the context in user comments, so as to overcome the problems of insufficient existing test data and untrue test data generated by existing methods; The invention makes full use of the user comments of speech recognition-related APPs, mines the contextual factors in speech recognition and the relationship between them, and extracts the values ​​of the contextual factors from the data set to build a test context classification tree model, thereby guiding the generation of speech test data. The quality assurance of speech recognition software in the process of frequent version evolution has important research significance and application value

Images