Comparative Study of LSTM and T5-Flan Models for Automated Test Case Generation from Gherkin Requirements

This paper investigates the application of LSTM and T5-Flan models for generating test cases from simplified Gherkin requirements in automated software testing. Automated test case generation using AI models has gained prominence recently, offering substantial benefits in efficiency and reliability. Therefore, understanding the comparative strengths of different models in this domain is helpful for optimizing software testing practices.In this study, we compare and analyze the performance of both models in terms of accuracy and efficiency, using a diverse dataset of Gherkin scenarios and evaluation metrics such as BLEU, ROGUE, BERTScore, and Semantic Similarity. The findings demonstrate that T5-Flan consistently outperforms LSTM across all metrics, showcasing superior capability in capturing complex dependencies and semantic nuances present in Gherkin requirements.