Pré-processamento de dados de texto de requisitos de software utilizando técnicas de aprendizado de máquina

Abstract

The execution of machine learning algorithms applying natural language processing has been helping humans and automating many everyday tasks. However, one of the most time-consuming steps in the implementation process is the treatment of dataset so that it is ideal for training the algorithm, which happens due to the impossibility of having a typing pattern or data structure, causing the texts available for searches to contain a lot of noisy data. Therefore, the present study is dedicated to the execution of techniques for pre-processing text data using regular expressions, to treat noises that have a writing pattern and machine learning techniques to seek solutions to replace noise in texts that they do not have a writing standard, in order to speed up the pre-processing of the data used for the implementation of natural language processing algorithms.