Question answering bot written in Ruby that learns through group chat interaction and context.
- Corpus of questions that she adds onto as she learns.
- Peri reads corpus to create a set of common tokens that are used in question detection.
- Corpus of "filler" data that cleans out junk tokens.
- This filler corpus is used to remove common tokens that aren't question specific.
- Individual corpora are created for each question containing a sample of answers.
- When asked a question, Peri replies from the sample.
- If the user is satisfied (gives a thumbs up), that answer is weighed heavier.
- Ideally, this system will allow her to more precisely answer questions.