CBL - Campus del Baix Llobregat

Projecte llegit

Títol: MyBotWilliam: Un ChatBot para una asignatura de programacion.


Estudiants que han llegit aquest projecte:


Director/a: MESEGUER PALLARÈS, ROC

Departament: DAC

Títol: MyBotWilliam: Un ChatBot para una asignatura de programacion.

Data inici oferta: 17-06-2020     Data finalització oferta: 17-02-2021



Estudis d'assignació del projecte:
    GR ENG SIST AEROESP
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
Inteligencia Artificial, Chatbot, web scraping
 
Descripció del contingut i pla d'activitats:
 
Overview (resum en anglès):
In this project we are going to develop a chatbot with artificial intelligence, with the aim to accelerate the resolution of doubts for a programming subject.
For the creation of the chatbot, we have created different datasets with questions and answers, extracted via web scraping from a well-known web page: Stack Overflow. To determine how the chatbot intends to find the answers, we have studied different selection methods: similarity, decision tree, neural nets and deep learning. A comparative of the performance of the different methods for each database is going to be done, with the objective to decide which model is best to apply to our chatbot. Finally, we are going to corroborate the correct functioning of the real-time chatbot.
After analysing different possible combinations of datasets and selection method technologies, the chatbot has been decided to build based on a neural net model with the unmodified dataset. This combination results in an accuracy efficiency of 80%. Nevertheless, even though its performance is good it does not achieve to accelerate the doubts resolution, because for the answer to be effective you have to know what questions are included in the database. Additionally, there is a possibility that the answer is related to the topic, because there is a word coincidence, but it is not really answering your question.
All in all, we can say that we have accomplished almost every objective of our project except the aforementioned answer acceleration. On the other hand, we have achieved to create different datasets and selection methods to develop a real-time chatbot.


© CBLTIC Campus del Baix Llobregat - UPC