Name: MARCOS ALÉCIO SPALENZA
Type: PhD thesis
Publication date: 27/04/2023
Advisor:

Name	Role
CLAUDINE SANTOS BADUE	Co-advisor *
ELIAS SILVA DE OLIVEIRA	Advisor *

Examining board:

Name	Role
CLAUDINE SANTOS BADUE	Internal Examiner *
ELIAS SILVA DE OLIVEIRA	Advisor *
PATRICK MARQUES CIARELLI	Internal Examiner *

Summary: The evaluation is a basic step that composes the learning assessment and guarantees progress
through the planned curriculum. On learning assessment, the application of open-ended questions
contributes to developing critical thinking and writing skills. The tutor must adapt your teaching methods at a large scale, avoiding assessment overload. Even though small documents, the answer collection produces
a large corpus. Meanwhile, the tutor should analyze details inside these answers to identify potential
learning gaps. Therefore, the adoption of educational supporting methods aims to improve the teacher’s
analytical capacity and, consequently, intensify monitoring of the student’s learning. Over these studies, we
present an Active Learning model for educational document classification, specifically for Short Answer
Grading problems. For this purpose, we combine clustering and classification models for pattern
recognition with grammatical, morphological, semantic, syntactic, statistical, and sequential analyses for
textual enrichment. The system aims to approximate the textual patterns to the tutors’ evaluation criteria,
reducing their effort and supporting the learning assessment. We apply our method to 65875 students’
answers among 255 questions found in Short Answers Graders’ literature. According to the human graders,
our proposal achieves an average accuracy of 79% and a weighted F1 of 78%.

Access to document