INFERMEDICA CHATBOT

Question Answering System For COVID-19 Questions Using NLP Techniques

Why INFERMEDICA?

The current situation worldwide pandemic COVID-19 since 2019.As a consequence, people have been looking for any available information about it ever since. So we resorted to creating a Chatbot to help get information from it,

• Chatbots are much more advanced now.
• Chatbot is one of the most interesting and important technologies.

Dataset (Question-Answer Pairs of Dataset about covid-19)

• The dataset consists of a total of 2,348 questions.
• With 280 unique Question IDs.
• It is annotated by a selected group of medical experts, scientists, engineers, technologists, and specialists.
• It relies on a repository of Frequently Asked Questions gathered from reliable sources such as the Centers for Disease Control and Prevention (CDC), the World Health Organization (WHO), the University of Washington Bothell, and the Federation of American Scientists.

Information Retrieval Approaches:

• First approach based on TF-IDF.
• Second approach based on LSA.
• Third approach based on Glove.

Concepts Serve The Three Approaches

1. Cosine Similarity.
2. Encoding (Representing text as numbers).

Neural Model Approach:

• sequence-to-sequence (seq2seq) model.

API with Flask

Challenges:

• Small dataset.
• Availability of GPU.

Future Works:

• Get larger dataset.
• Next Approaches
• Model supports other languages.

Eventually, we hope that our project can provide some basics line in implementing a good question answering chatbot and what comes next could be more.