Desenvolvimento de um sistema de tradução automática adaptado ao domínio de documentações técnicas de tecnologia

Abstract

Automatic translation uses computational methods to convert texts between different languages. Domain adaptation is essential for improving the accuracy of translations in specific domains. This work aims to develop an automatic translation system specialized in technology technical documentation. To achieve this, the fine-tuning of an existing model will be conducted using texts specific to this domain. Subsequently, an API and Graphical user interfaces (GUIs) will be created for end-users to access the adapted model. The fine-tuning will be performed using the Transformers library on a machine with two NVIDIA T4 GPUs and 30GB of RAM. For the development of the GUIs, will be employed HTML, CSS, JS, and the React framework. The results will be presented in terms of metrics and comparisons of translations both before and after the fine-tuning, enabling a more comprehensive evaluation of the model’s performance.