The Finalist | QA Chatbot
Our idea was to create Question Answering Chatbot capable of talking about safety during traveling. We implemented a AI model based on BERT (Natural Language Processing state-of-the-art model from Google Research – https://github.com/google-research/bert ). Model was trained on web crawled data about the COVID-19 situation. Model is English speaking because English is an international language and English BERT is more popular.
We started from gathering the data from the websites. Data for example from polish websites was in Polish so we used Google translator API to translate the documents. The translated data was feed into the model and the final model was used to create a demo website where user’s can ask question about the COVID-19.
We fully developed backend, frotend and AI model.
Improvement of collected data and its extension. Due the short amount of time we were not able to translate all data. The model was trained on small portion of it and would perform much better when trained on bigger dataset. Better selection and processing of data would also improve the predictions.
Team name: Dziobak
Project name: QA Chatbot
Team members: Jakub Kowalski, Dominika Bankiewicz