Generación de código a partir de requisitos basados en voz. Un análisis multicriterio de herramientas de generación de código basadas en Inteligencia artificial

Abstract

This research entitled: “Code Generation from Voice-Based Requirements. A multi-criteria analysis of artificial Intelligence-based code generating tools” focuses on selecting, evaluating, and comparing artificial intelligence tools applied to a case study on the registration and management of pets in the city of Cuenca, for the generation of code through voice-based prompts, where they are applied patterns in the inputs provided to obtain more complete results. An exhaustive investigation of the types of existing tools is carried out, in addition to interviews with professionals in the software area where a set of 10 tools are analyzed; The selection criteria are that they generate code, that the tool accepts voice input dictated or cloned, and that it has a free use license. As a result, 4 AI tools are obtained that meet these requirements: ChatGPT, Bing, Gemini and Replit, to which a review of use cases is applied through tests carried out nearly with the prepared prompts, also performing a multi-criteria analysis based on Interface features, voice accuracy, ease of use, functions and compatibility through manual calculation processes and supported by the RStudio tool. Finally, a guide to these 4 resulting tools is provided.