Voice recognition system crowdsourcing test case generation method based on test requirements

Abstract

The invention discloses a voice recognition system crowdsourcing test case generation method based on test requirements. The method is characterized in that a provided voice recognition system is tested based on a crowdsourcing technology, the test requirements submitted by a task requester are recognized and extracted, a voice recognition keyword list in a preset keyword library is used for filtering out texts related to test case content from recognized text information fragments, and the texts serve as candidate test requirements. Three parts are included: a voice recognition test parametercollection module, a text extraction module and a test case generation module. Input is a text test demand, feature extraction is performed based on a user demand, and a test template is generated. The method has the following beneficial effect that the multi-dimensional and diversified requirements of ASR testing are met. The test participation threshold of crowdsourcing workers is reduced, testtasks are not limited by places and identities, and the crowdsourcing workers can sign up to participate in the test as long as the crowdsourcing workers have test tools and test capabilities, so that the test loss is reduced, and the product research and development period is shortened.

Description

technical field [0001] The invention belongs to the field of software testing, in particular to crowdsourcing testing and generation of test cases. The present invention tests the provided speech recognition system based on crowdsourcing technology. Before calling crowdsourcing workers, it is necessary to generate test cases based on the test requirements provided by task requesters, and perform crowdsourcing task assignment around the test cases. Background technique [0002] The implementation of artificial intelligence software, such as speech recognition (ASR) systems, cannot do without a large amount of audio data. Whether it is verifying the robustness of the system at the data ingestion layer or verifying the accuracy of the system at the core algorithm layer, a large amount of data is also required. In the ASR training phase, it is difficult to complete the collection and cleaning of large-scale data with high quality; in the ASR evaluation phase, the quantity and q...

AI Technical Summary

Problems solved by technology

The test steps and methods are very simple and understandable for professional testers, but for those who have not been in touch with the speech recognition system, they lack relevant background knowledge and have no way to start, resulting in the inability to complete the test task on time

Images