TextFlowReduce: utilizando threads e expressões regulares para análise de questões

Abstract

The purpose of this paper is to present the development and validation of the TextFlowReduce library, an automated solution for analyzing short essay responses based on configurable semantic criteria and parallel processing. The library was designed to reduce subjectivity and manual correction time, allowing teachers to define criteria, which should be formulated using keywords, phrases, and weights, depending on the methodology and criteria chosen. The methodology adopted in this work involved implementation in C, using Regular Expressions (Regex) for accurate pattern searching and the Task Parallel Library (TPL) for concurrent analysis of the criteria. The experiment was conducted with 30 participants and demonstrated that the tool is capable of objectively quantifying technical knowledge, providing detailed performance reports and identifying conceptual gaps. The results showed a strong correlation between the use of technical terms and the final score, validating the deterministic and transparent approach. Thus, it is concluded that the TextFlowReduce tool is effective for automating the evaluation of discursive responses, optimizing the evaluation process and providing immediate and detailed feedback.