Citation Parsing

Previous Research

• A New Approach towards Bibliographic Reference Identification, Parsing and Inline Citation Matching
• A structural SVM approach for reference parsing
• ParsCit: An open-source CRF reference string parsing package

Possible Approaches

• Stochastic Context-Free Grammars -- Haven't been applied to this problem yet (as far as I know).
• Conditional Random Fields -- This is the approach of ParsCit, FreeCite, etc. They tend to have about 80% accuracy.
• Crowd Sourcing Parsing -- Using a game or Mechancial Turk like system, parse and tag some citations.
• Programming Contests -- Once we have a large data set of outputs and inputs, we can start asking for programmers to solve the problem in the broader internet community.

Lessons Learned So Far

• Parsley will occasionally get stuck in infinite loops when it parses a null string.
• Appropriate git branching is necessary to keep main clean and collaborate.
• Parsley Grammars can call arbitrary python code, this hasn't fully been taken advantage of yet.