Por favor, use este identificador para citar o enlazar este 铆tem:
https://dspace.unl.edu.ec/jspui/handle/123456789/30620| T铆tulo : | Modelo QA basado en DistilBERT para responder a preguntas sobre el contenido extra铆do de tareas acad茅micas de la carrera de Computaci贸n de la UNL. |
| Otros t铆tulos : | QA model based on DistilBERT to answer questions about content extracted from academic assignments of the Computer Science course at UNL |
| Autor : | Cumbicus Pineda, Oscar Miguel Jim茅nez Merino, Edy Francisco |
| Palabras clave : | MODELO QA DISTILBERT DATASET SQUAD1.0 CRISP-ML(Q) ROUGE |
| Fecha de publicaci贸n : | 20-sep-2024 |
| Editorial : | Universidad Nacional de Loja |
| Resumen : | La adaptaci贸n de modelos pre-entrenados Question Answering (QA) es una tarea esencial para que estos puedan ser implementados en diferentes escenarios. El objetivo de esta investigaci贸n es obtener el valor de la m茅trica ROUGE al aplicar la t茅cnica Fine-Tuning sobre el modelo DistilBERT para dar respuesta a preguntas sobre el contenido extra铆do de tareas acad茅micas de la Carrera de Computaci贸n de la Universidad Nacional de Loja. Para desarrollar este trabajo se us贸 la metodolog铆a CRISP-ML(Q) como marco de referencia, haciendo uso de sus cuatro primeras fases, en las que se realiz贸: una recopilaci贸n de 30 tareas acad茅micas obtenidas de 6 materias diferentes, de las que se gener贸 80 preguntas sobre su contenido a trav茅s de crowdsourcing, las cuales sirvieron como base para crear un dataset en formato SQuAD1.0 con 1410 datos, de los cuales 800 se generaron mediante par谩frasis y el enfoque Few-shot learning, y los 610 restantes con el aporte directo del autor, este dataset se dividi贸 en 90% para entrenamiento (train) y 10% para evaluaci贸n (test), con una subdivisi贸n adicional del conjunto train (75% train y 25% validation), teniendo los datos preparados se ajust贸 hiperpar谩metros de DistilBERT para entrenar cuatro modelos diferentes usando TensorFlow en la plataforma Google Colab con el entorno de ejecuci贸n GPU T4, seleccionando el mejor modelo en base a su nivel de extracci贸n de respuestas y F1-score. Una vez elegido el modelo QA, se realiz贸 una evaluaci贸n mediante la m茅trica ROUGE incluida una prueba A/B testing. El modelo QA se despleg贸 en Hugging Face y logr贸 una precisi贸n de 86,93% durante su entrenamiento con 51 茅pocas, learning_rate de 1e^(-5) y batch_size de 32, el cual mediante la evaluaci贸n logr贸 un F-measure m谩ximo en ROUGE-L de 60,96. Estos valores demuestran la importancia de aplicar el Fine-Tuning en el desarrollo de modelos QA para contextos espec铆ficos. Palabras Clave: modelo QA, DistilBERT, dataset SQuAD1.0, CRISP-ML(Q), ROUGE |
| Descripci贸n : | The adaptation of pre-trained question-answering (QA) models is an essential task so that they can be implemented in different scenarios. The objective of this research is to obtain the value of the rough metric by applying the Fine-Tuning technique to the DistilBERT model to answer questions about the content extracted from academic tasks of the Computer Science Department of the National University of Loja. To develop this work, the CRISP-ML (Q) methodology was used as a reference framework, making use of its first four phases, in which the following was done: a compilation of 30 academic tasks obtained from 6 different subjects, from which 80 questions about their content were generated through crowdsourcing, which served as a basis for creating a dataset in SQuAD1.0 format with 1410 data, of which 800 were generated through paraphrasing and the Few-shot learning approach, and the remaining 610 with the direct contribution of the author. This dataset was divided into 90% for training (train) and 10% for evaluation (test), with an additional subdivision of the train set (75% for train and 25% for validation). Having the data prepared, DistilBERT hyperparameters were adjusted to train four different models using TensorFlow on the Google Colab platform with the GPU T4 runtime environment, selecting the best model based on its level of response extraction and F1-score. Once the QA model was chosen, an evaluation was performed using the ROUGE metric, including A/B testing. The QA model was deployed in Hugging Face and achieved an accuracy of 86.93% during its training with 51 epochs, a learning rate of 1饾憭 -5, and a batch size of 32, which through evaluation achieved a maximum F-measure in ROUGE-L of 60.96. These values demonstrate the importance of applying Fine-Tuning in the development of QA models for specific contexts. Keywords: QA model, DistilBERT, SQuAD 1.0 dataset, CRISP-ML(Q), ROUGE |
| URI : | https://dspace.unl.edu.ec/jspui/handle/123456789/30620 |
| Aparece en las colecciones: | TRABAJOS DE TITULACION AEIRNNR |
Ficheros en este 铆tem:
| Fichero | Descripci贸n | Tama帽o | Formato | |
|---|---|---|---|---|
| EdyFrancisco_JimenezMerino.pdf | 4,26 MB | Adobe PDF | Visualizar/Abrir |
Los 铆tems de DSpace est谩n protegidos por copyright, con todos los derechos reservados, a menos que se indique lo contrario.