AI Zone Admin Forum Add your forum

NEWS: survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

Self-supervised learning

I’m still defining my bot project, the characteristics that I have thought of supervised learning, and the information obtained may not be entirely trusted.

Supervised learning can be a laborious and time consuming, moreover, that by not focusing on a particular topic of conversation and be varied, some of them certainly not the domain.

So I thought of a supervised learning more or less automatic. At the end would not supervised by me, but through their own subjects.

How I will do this? Using a scoring system for each user, it is the user who will make introductory information and to validate the information entered by other users.

When someone checks the previously entered information that other users are few points of trust and therefore the rest of the information provided becomes more valuable.

However, if the information that a user is questioned by other users original user loses trust points and could even do that any information entered by him be removed from the knowledge base.

I hope I explained correctly, and you tell me what you think of this learning system.


  [ # 1 ]

  I am most familiar with the ultra Hal personal assistant from Zabaware which is a self learning bot, at this time Zabaware has put this bot at one or more websites and just collecting data from the conversations the bot will be engaging in. At this time I am not sure how the data will be selected or rejected for use.
  At this time I am currently trying to put together an Alice type bot that will be self learning, again this is for personal use and I would just like to add the capability of learning rather than being able to accept or reject information.  If your bot is to be placed on a website more than likely you will definitely need an automated system.


  login or register to react
‹‹ Debugging AIML?      Google Puzzles ››