I'm going through this chatbot example, which uses the Cornell movie dialog corpus. Expanding this example, is it possible to build a "knowledge bot" (ie) a bot that can chat and be knowledgeable in a specific domain? So for example, I use the Bible to train the system. Then it should be able to answer questions on Bible. Or I use the wikia content of Star Wars and be able to answer all the questions related to Star Wars. Is such a system even possible, or is it still in the realm of scifi?
$\begingroup$
$\endgroup$
2
-
1$\begingroup$ I guess IBM' Watson kind of does this, answering questions. Take a look at this for a demo $\endgroup$– Kiritee GakCommented Jan 31, 2017 at 10:36
-
$\begingroup$ I would like to have a bot analyze/deep learn the Bible and then use that knowledge to cross check:reference other data sets to generate similar words comparisons or write a new book maybe a modern bible or a modern theory on life and death or something. . The other data sets could be 10000 articles from various news articles, or 10,000 best sellers books, or etc. I’m not familiar with the jargon here -but I would pay to have someone help me create This bot or tool. A. Learn/analyze the Bible - 1. Analyze 10k best seller books/published studies (self help, scientific, astronomy categories) 2. $\endgroup$– Chelsea SirenoCommented Jun 14, 2018 at 6:11
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
It is definitely possible, but obviously to a certain extent. What you need, is a sequence to sequence model trained on questions and answers data of a domain. Denny Britz has this amazing blog post on impelementing a retreival based chatbot trained on ubuntu dialog corpus using tensorflow. Go through the post to understand the difference between retreival based and generative model chatbot. The efficiency and efficacy of such a bot, is another question all together and rigorous research is going on to improve such bots. IBM watson does it pretty well as mentioned in the comments.