27 Jan 2013

This was my attempt at combining two openFrameworks addons: ofxEliza and ofxSpeech.  The goal was to create an implementation of the historic keyword-based Eliza chatbot that could utilize ofxSpeech to both recognize audible keywords and respond using synthesized speech.  Both addons successfully compiled together, but the Eliza module seems to have some issues, as demonstrated in the video.  Namely, the chatbot is great at detecting edge cases like repetitions and short responses, but doesn’t actually pick up any keywords, even when typed into the console.  This doesn’t make for a great therapist.  I spent time trying to debug the input parser for Eliza, but didn’t make much progress and as a result I didn’t dive deep into speech recognition.  An alternative to ofxSpeech is ofxGSTT, which uses Google’s speech to text engine but is more complicated and requires the integration of additional addons.  Eliza’s keyword-based responses should match well with ofxSpeech’s dictionary-based recognition.

The OF code can be found here.

Screen Shot 2013-01-28 at 12.20.49 AM