-
Notifications
You must be signed in to change notification settings - Fork 6
Home
micro222 edited this page Oct 30, 2014
·
10 revisions
I've been working on a chatbot on and off for about 11 years, and while I've made a lot of progress, there's a great deal of work left to do. I was thinking that if this was a group project, development would pick up speed considerably, plus it could be more fun to work on. I'd like see if there's any interest out there.
I wanted to make a chatbot that is better than the AIML types. Something that would pay attention to who it was talking to and try to learn from that person. It would not guess at an answer when it has no understanding of that was said. If it doesn't understand, it will explain why and ask for information in an attempt to understand. What it learns can be applied later in the conversation and with other users.
This is a text based robot with the learning as the main emphasis
It tends to try to learn about the people it talks to and the world in general
Its database relies on a classification of basic knowledge
It uses templates to interpret sentances
Only certain types of words are allowed to fit in the templates
It keeps track of the people that talk to it
It starts out with simple features and is expandable
ID numbers are used identify specific members of a category, usualy people could also be pets, significant rocks, buildings db entries start with a number sign followed by the id number
Will expect to know who it's talking to
Will remember elements of conversations
Will form opinions of the people it talks to
Will try to use people to expand it's knowledge
Will try to make sense of every part of a sentence and will reject any word it can't make sense of.
Will react to things that are odd
Only full sentences
No pronouns
No figures of speech
No slang
No compound sentences
Ignores multiple meanings