Adviser Found!
Yesterday I spoke with Dr. Charles Romney about the Personified Chat Bot. He seemed very interested and agreed to be my MSIS Project adviser. The next step is to officially list him as my adviser in the UA Little Rock system. I’ll also need to assemble 2 more faculty members, forming a committee of 3.
Project Ideas
Dr. Romney and I discussed some potential paths for this Personified Chat Bot. The basic idea is to unify the features of a Markov chain and a natural language generator.
Think Twitter bot version of Donald Trump
Given a corpus of text from a given speaker/writer/tweeter, this chat bot could generate text both according to natural English and according to the author’s stylistic autograph. So maybe Chat Bot is a misnomer. Maybe this is more of a Twitter-targeted bot: Personified TweetBot.
One approach would use a natural language parser to bin the entire corpus into parts-of-speech, creating a repository of potential subjects, verbs, adjectives, etc. In parallel, a Markov chain algorithm could be assigning all the transition properties for the chosen n-gram of words. The part-of-speech “raw material” could then be utilized by a natural language generator to create many possible outputs, and the markov transition properties could determine the single output by selecting the alternative with the best fit to the corpus distribution.
Research Items
- Natural Language Generation (NLG)
- Wikipedia: https://en.wikipedia.org/wiki/Natural_language_generation
- Python NLG project from toumorokoshi: https://github.com/toumorokoshi/nlgen
- Natural Language Parser
- Parse theory intro and research project: https://explosion.ai/blog/parsing-english-in-python
- Stanford Parser: https://nlp.stanford.edu/software/lex-parser.shtml
- UPDATE 11 Sep 2017 - Association for Computational Linguistics (ACL) Open Source Literature