Stephen Bly

Ryhan Hasan

Daniel Sedra

Team 6

We used Python (mostly NLTK) and Java (arkref). We are still working on coreference resolution.

Ask:

First splits file into sentences and then ranks them based on length and presence of 'It is' or 'is'. It then generates question using handwritten rules.

Example:

./ask Question_Answer_Dataset_v1.1/S08/data/set1/a1.txt 4

generates 4 questions from a1.txt

Answer:

Given a quesition we search for relevant sentences using n-gram overlap and keyword proximity. This sentence is then returned.

Example:

./answer elephant

goes to Question_Answer_Dataset_v1.1 and each year answering every question from every year.