An information retrieval system built to answer any question written in plain English by returning the most relevant exerpt from a library of text.
Final product employs a number of natural languages processing techniques, including
- Word dependency parsing
- Word lemmatization
- Pronoun resolution
- Part of speech tagging
- Synset matching
- Question classification
- Discourse model for creating pronoun-to-proper-noun associations
- Fitness scoring for candidate answers