Leveraging Large Language Models for Automated Web-Form-Test Generation: An Empirical Study

The testing of web forms is an essential activity for ensuring the quality of web applications, which mainly involves evaluating the interactions between users and forms. Automated test-case generation remains a challenge for web-form testing: Due to the complex, multi-level structure of web pages, it can be difficult to automatically capture their inherent contextual information for inclusion in the tests. Large Language Models (LLMs) have great potential for contextual text generation. OpenAI's GPT LLMs have been receiving a lot of attention in software testing, however, they may fail to be applied in practice because of information security concerns. To the best of our knowledge, no comparative study examining different LLMs has yet been reported for web-form-test generation. To address this gap in the literature, we conducted a comprehensive empirical study investigating the effectiveness of 11 LLMs on 146 web forms from 30 open-source Java web applications. According to the experimental results, different LLMs can achieve different testing effectiveness. Notably, the GPT-4, GLM-4, and Baichuan2 LLMs can generate better web-form tests than the others. Compared with GPT-4, other LLMs find it difficult to generate appropriate tests for web forms, resulting in decreased successfully-submitted rates (SSRs, measured by the proportions of the LLMs-generated web-form tests that can be successfully inserted into the web forms and submitted) ranging from 9.10% to 74.15%. Nevertheless, some LLMs achieve higher SSRs than GPT-3.5, indicating a better ability to generate appropriate tests for web forms. Our findings also show that, for all LLMs, when the designed prompts include complete and clear contextual information about the web forms, more effective web-form tests were generated. Finally, we offer some insights for using LLMs to guide automated web-form testing.