Speech Application Test Script Generation via Voice Call Traversal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech application testing methods are slow, error-prone, and require substantial human effort, struggling to simulate realistic human speech and handle multiple languages, with automated test scripts being time-consuming and expensive to maintain.
Innovation Solution
A computer-implemented method and apparatus that generate test scripts by initiating voice call interactions with speech applications, traversing interaction nodes, and providing responses, allowing for automated testing with minimal human intervention and optimal coverage of interaction nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated test scripts are used to increase testing speed, then productivity improves, but device complexity and maintenance cost increase
Solution Approach 1:
The system performs self-testing by automatically generating test scripts from its own speech recognition and synthesis operations. The speech application under test generates test data through normal operation, eliminating the need for external test script authors and reducing maintenance burden.
Solution Approach 2:
Instead of creating test scripts to test the speech application, the system inverts the approach by using the speech application's own output and behavior to generate the test scripts. The application under test becomes the test generator, reversing the traditional testing paradigm.
2Measurement precision
If comprehensive test coverage is achieved by manually analyzing complex grammars and cross-referencing documents, then measurement precision improves, but loss of time increases
Solution Approach 1:
The system pre-generates comprehensive test scripts by traversing the grammar structure and identifying all possible utterance paths before actual testing begins. This preliminary generation ensures complete coverage without time-consuming manual analysis during execution.
Solution Approach 2:
The system replaces manual mechanical analysis of complex grammars with automated computational traversal of the grammar structure. The computer systematically explores all possible paths through the grammar, substituting human effort with algorithmic processing.
3Reliability
If test scripts are regularly updated to match speech application changes, then reliability improves, but loss of time and productivity decrease
Solution Approach 1:
The system automatically updates its own test scripts when the speech application changes. By monitoring modifications to the application and regenerating test scripts from the updated structure, the system maintains reliability without requiring external intervention or time-consuming manual updates.
Solution Approach 2:
The system implements feedback loops where test execution results and application change detection trigger automatic regeneration of test scripts. This feedback mechanism ensures test scripts remain synchronized with the application state without manual intervention.
4Measurement precision
If multiple languages and accents are tested with realistic human speech simulation, then measurement precision improves, but device complexity increases
Solution Approach 1:
The system uses a universal speech generation mechanism that can produce multiple languages and accents through a single platform. The speech application generates test utterances in various languages and accents, eliminating the need for separate testing systems for each language variant.
Data Source
AI summary
A computer-implemented method and an apparatus for facilitating speech application testing generate a plurality of test scripts. A test script is generated by initiating a voice call interaction with a speech application including a network of interaction nodes, and repeatedly performing, until a stopping condition is encountered, the steps of, executing the voice call interaction by traversing through interaction nodes until an interaction node requiring a response is encountered, selecting an utterance generation mode, determining a response to be provided corresponding to the interaction node, and providing the response to the speech application. The test script comprises instructions for traversing interaction nodes and for provisioning one or more responses during the course of the voice call interaction. One or more test scripts from among the plurality of test scripts are identified based on a pre-determined objective and provided to a user for facilitating testing of the speech application.


