Name: RODRIGO RUY BOGUSKI
Publication date: 05/12/2024
Examining board:
Name |
Role |
|---|---|
| ALBERTO NOGUEIRA DE CASTRO JUNIOR | Examinador Externo |
| CAMILA ZACCHÉ DE AGUIAR | Coorientador |
| CREDINÉ SILVA DE MENEZES | Examinador Externo |
| DAVIDSON CURY | Presidente |
| MONALESSA PERINI BARCELLOS | Examinador Interno |
Pages
Summary: People grouping can be applied in various everyday activities and in different
contexts. It can be done in different ways and become a highly complex task
depending on a series of criteria. One relevant criterion is comparison based on
cognitive semantic similarity, as it allows identifying groups based on shared
knowledge or common meanings. This approach can be particularly useful in
contexts where people can be grouped based on their knowledge of a subject,
aiming to facilitate collaboration and cooperation in the construction and
exchange of knowledge.
An effective way to organize and visualize this shared knowledge is through
conceptual maps. Concept maps are powerful visual tools for representing and
organizing knowledge in a clear and structured manner, helping to illustrate the
relationships between different concepts and ideas, thus facilitating
understanding and learning. The use of these maps is essential, as it allows us to
visualize how information connects, as well as helping to identify knowledge gaps
and encouraging reflection on complex topics. In this way, by using conceptual
maps, it becomes possible to foster a more collaborative and structured approach
to information exchange, problem-solving, and knowledge representation.
In this context, this research proposes a framework that enables grouping and
analysis based on the semantic similarity of conceptual maps. The framework
consists of the following modules: 1) automated reading of conceptual maps, with
the representation of information obtained from the map in the form of a
descriptive text; 2) text normalization to make it suitable for processing by
natural language semantic vector models; 3) thematic analysis and grouping
using machine learning techniques that consider the thematic context; 4)
semantic analysis and grouping using neural networks trained to identify the
semantic context; 5) quality analysis of the groups from different perspectives
using data mining association rules.
To instantiate the framework, we developed software whose conceptual model
implements the framework components, enabling its application in teaching-
learning activities. We used the software in different classroom experiments, with
three groups from two educational institutions. The results of the experiments, in
this context, show a match of over 80% between the semantic groups proposed
by the teachers and those generated by the framework, thus confirming its
relevance for performing and understanding semantic groupings in different
scenarios where knowledge can be produced by teams.
