The dream

The plan

• Scrape Wikipedia
• Use NLP to find good questions
• Use WordNet to create good filler lies
• Build a web app to serve it up well

The result

• Scrape Wikipedia
• Use NLP to find poor questions
• Use WordNet to create bad filler lies
• Build a web app to serve it up well enough

NLP

• Getting Wikipedia data was easy
• Identifying parts of speech was good
• Coming up with a good algorithm was hard:
  • Try to blank out unusual words
  • Try to blank out proper vs. improper nouns
  • Try to create a classifier

https://github.com/atbaker/wikipedia-question-generator

Web app

• MEAN stack
• Socket.io + Angular-ui-router is awesome
• Bootstrap isn't great for games
• It works... mostly...

http://wikitrivia.atbaker.me

DEMO

Our articles:

Tony Bennett

Python (programming language)

Henry V, Duke of Carinthia

Scabbling

Globe of the Great Southwest

Ukrainian Women's Volleyball Super League

Final thoughts

• NLP was fun to learn
• Unexpected characteristics of a "good" question
• Socket.io + Angular-ui-router is awesome
• Question generating code on GitHub