Week 21: Advancing with the implementation

The team aimed for a continuity of the work done last week, namely improving the train of the BERT model to be more suitable to our Energy work theme. We continued to improve our dataset generation both in quality as in quantity, to increase the accuracy of intent and BIO classification. We also started to implement our chatbot so that it will be able to correctly identify the type of text/input from the user and give an appropriate response, transmitting the sensation that it was responded by a human, as much as possible.

On the classes of AAUT2IA, the focus was on Deep Learning techniques, with special focus on Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN). On the other hand, on LNSCIA, the analyses of sentences using Sent2Vec or StarSpace models were taught, as well as Topic Modelling and Machine Translation techniques.

On a more personal note, it was a rough and more challenging week than usual, as the team had to face with Bruno Veiga surgery recovery and Miguel Silva acute sinusites, which, at some times, left the team with half of its members to advance with the project implementation.