- Protobowl training data
- Compressed CSV file
- id Protobowl identifier of the question
- sent Sentence id for the sentence (goes from 0 to N within a single question id)
- text The text of the question (i.e., the question itself) for this sentence (a single question consists of multiple sentences)
- ans The human-written answer line of the question
- page The corresponding Wikipedia page for the answer line (some are blank)
- fold The data fold that we use internally
- Streamlined system (encouraged for baseline system; available August 2017)
- Complete system (code is messy and under active development)
- Server-side code (to run your own server for testing)
- Simple client code
Feel free to ask questions about these resources on our mailing list.