Team Met on Piazza's solution for Carnegie Mellon's 11-411 Natural Language Processing class project. The project is a Wikipedia question answering and generation system. Given a Wikipedia article, the system both answers questions about the content of the article and generates new questions about the article. We ended up placing 2nd in the class competition.
To learn more about how our system works, watch this video.
Checkout the repo and cd to the root.
Then install the requirements by running
pip install scripts/question_answering/qa_requirements.txt
Once you have the requirements installed, run the script interactively by entering
python3 -i scripts/question_answering/answer_question.py [set_num] [doc_num]
To test a question, enter
get_best_sent(<question_text>)
To test questions about article 1 from set 1 (Old Kingdom), enter
python3 -i scripts/question_answering/answer_question.py 1 1
Then, to test the question "Who is King Djoser?", enter
get_best_sent("Who is King Djoser?")
Note: a number of sample questions are included for the very first article (set 1 article 1), and are stored in the array S1A1.