Demo web server app that shows how BERT model trained on SQuAD dataset deals with the machine comprehension task.

It was implemented for the purposes of presentation on STX Next Tech Power Summit 2019. Slides from the presentation are available here.

Before running the app, please make sure that you have all requirements installed and download BERT model for machine comprehension.

To run the server, navigate into the main project's directory and type

$ flask run

Requests sent to the BERT web server should have the following form:

{
  "context": "Bert is a yellow Muppet character...",
  "question": "Who was the first Bert performer?"
}  

You can expect answers of the form:

{
    "answer": "Frank Oz"
}