Sporala red del conocimiento
EN
Universidad Nacional de Colombia

Bogotá - Ingeniería - Maestría en Ingeniería - Ingeniería de Sistemas y Computación · 2023

Development of a software method to assist in the thematic analysis of responses to open ended questions in Spanish-language surveys

Thematic analysis is fundamental in qualitative research, providing rich insights but often requiring substantial time and expertise. This work addresses some limitations of existing Computer-Assisted Qualitative Data Analysis Software (CAQDAS) and presents a novel method specifically designed to assist in the thematic analysis of multi-label open-ended questions in Spanish-language surveys. The proposed method melds domain expertise with advanced language models to establish preliminary categories. Subsequently, human discernment is combined with similarity measures to streamline the categorization of some responses using these preliminary categories. The process culminates in a robust and scalable automated categorization, utilizing diverse models, language models, and accuracy metrics. The proposed method is composed of three modular phases that can function independently or collaboratively, offering a comprehensive solution for researchers. It can reduce the labor-intensive coding process by leveraging Large Language Models (LLMs) and Natural Language Processing (NLP) techniques. The method's efficacy is evaluated through its application on a dataset from the National University of Colombia, demonstrating promising results across its various modules and pathways. The work opens avenues for further research, particularly in enhancing qualitative analysis methods with the integration of modern tools. (Texto tomado de la fuente)

Texto completo 72 páginas con texto de 76

Leer la tesis completa Ficha en el repositorio

Tesauro de su biblioteca