Chatbot para resolver inquietudes académicas sobre estudios de posgrado en la Carrera de Sistemas/Computación de la UNL
In the Systems/Computer Engineering career at the National University of Loja (UNL), students after completing their third level studies often choose to continue their professional training with postgraduate programs. However, obtaining immediate information about it is tedious due to the long respo...
Kaydedildi:
| Yazar: | |
|---|---|
| Materyal Türü: | bachelorThesis |
| Dil: | spa |
| Baskı/Yayın Bilgisi: |
2023
|
| Konular: | |
| Online Erişim: | https://dspace.unl.edu.ec/jspui/handle/123456789/27656 |
| Etiketler: |
Etiketle
Etiket eklenmemiş, İlk siz ekleyin!
|
Etiketle | Özet: | In the Systems/Computer Engineering career at the National University of Loja (UNL), students after completing their third level studies often choose to continue their professional training with postgraduate programs. However, obtaining immediate information about it is tedious due to the long response times given by the administrative staff through email or social networks of the UNL. For such reason the following research question arises, "Will a Chatbot implemented with the BERT language model, help students and professionals to have an answer in front of the concerns about the postgraduate processes that occur in the Systems/Computing Career of the UNL?". In this context, the following objective was set, "To implement a Chatbot for the UNL Systems/Computing Career, using the BERT language model, which allows to solve the doubts of students and professionals about the processes that are governed before and during the postgraduate studies". The project was completed using the methodologies of Barbara Kitchenham and Extreme Programming (XP). The former allowed for a systematic literature review (SLR) to understand the functioning of the BERT model and the technologies needed for its training and adjustment. The second was used in the development of the web interface using the planning, design, coding and testing phases. During this process, a dataset was created with information on postgraduate studies, which was used to train the model (a version called DistilBERT was chosen), obtaining an accuracy of 89.2%. Subsequently, the chatbot architecture, programming and a series of tests were designed in an environment with students and professionals that allowed evaluating through a survey the acceptance of the virtual agent, obtaining a result of 83.9%, which affirmed that the chatbot was able to answer "all" the questions posed by the interested parties. Thus, it was concluded that the development of the virtual assistant using the BERT model can help students and professionals to a greater extent to solve their doubts regarding fourth level studies. Keywords: Conversational Agent, DistilBERT, BERT, Artificial Intelligence |
|---|