Mobile application public test report consistency detection method based on image text semantic fusion analysis

The invention discloses a mobile application public test report consistency detection method based on image text semantic fusion analysis. A text enhancement module, a classifier module, a decomposer module and a detector template are included. The text enhancer module enhances the text description so that test reports with different types of defects are uniformly distributed. The classifier module firstly constructs a classification method for defects in the crowdsourcing test report. And then a classifier is constructed based on a BERT model. After the classifier classifies the defects, a decomposer is introduced to analyze screenshots and text descriptions of the application program. The detector can detect the consistency of the crowdsourcing test report according to different strategies through the features extracted from the text description and the control containing the additional information extracted from the application program screenshot. 一种基于图像文本语意融合分析的移动应用众测报告一致性检测的方法，包括文本增强模块、分类器模块、分解器模块和检测器模板。