Chatbots are not something new: I encountered them online since 2004, when I started graduate school. You can find some of them under this link: https://sites.google.com/site/webtoolsbox/bots . There is another one called CleverBot: http://www.cleverbot.com/.
However, you can see that they are pretty dumb when you try them out. I have a feeling that those conversational models are Markov models, and the chatbots basically forget what they previously said.
A few days ago, Google put a paper on arXiv preprint that described how they applied neural network conversational models in chatbots. [Vinyals & Le 2015] In short they look for more previous sentences in the conversations. It is certainly inspired by some previous natural language processing (NLP) work from Google, especially Word2Vec that employs skip-gram models. [Mikolov et. al. 2013]
Deep learning is what a lot of big guys are trying now.
I wish to play with this chatbot.
- Web Tools for Kids (Chatbots): https://sites.google.com/site/webtoolsbox/bots
- CleverBot: http://www.cleverbot.com/
- T. Mikolov, I. Sutskever, K. Chen, G. Corrado, J. Dean, “Distributed Representations of Words and Phrases and their Compositionality,” Advances in Neural Information Processing Systems 26, pp. 3111-3119 (2013). [arXiv:1310.4546]
- Michael Czerny, “Modern Methods for Sentiment Analysis,” District Data labs (2015).
- O. Vinyals, Q. Le, “A Neural Conversational Model,” arXiv:1506.05869 (2015).
- N. McAlone, “Google’s artificial-intelligence bot says the purpose of living is ‘to live forever’,” Business Insider, Jun 27, 2015.