Enhancing AIML Bots using semantic web technologies
A little bit of technical difficulties…so Eric is explaining how he tried to explain his job to his Mom when he first started in SGML. I feel his pain…
“You know the Star Trek computer? We are building the data for the Star Trek Computer, RIGHT NOW!” That won’t work for me ’cause my Mom doesn’t like Star Trek…heck she probably thinks that is the one with R2-D2.
Ok, so Eric is still having technical trouble but he’s forging ahead anyway. This bot is based on the Elisa setup for AI (you type something and the computer responds). The difference is that A.L.I.C.E/AIML allows you to teach the bot anything you want…by adding XML (I assume), I think we’ll get there.
<that> tag is what adds context (ALICE remembers what she last said) – “is it raining outside” “yes” “then get an umbrella”
What drew him to RDF is the subject/predicate/object and the vast amount of RDF data sets available online….Linking Open Data project
Also included WordNet because it allows the bot to start learning the lexical structure of the language (dog=noun or verb, what kind of dog, specific dog?) That is cool. I need to look into Word Net some more. He is using FOAF but I wonder if any of the social networking sites would work? Just looking around the room at laptops in front of me I see a couple linkedIn users in almost every session….which makes sense since it is focused on business and FOAF is more esoteric…although cool and something else I will be checking out.
“Star Trek is 300 years from now so we’ve got time.”
People don’t talk like they type and people don’t type like they talk. “Speech recognition sucks. It’s getting better, but it isn’t there.” The interaction part with AIs is the weak link. Verbal interaction is much different than typing.
Seriously, check out Uche Ogbuji’s comments on this stuff. He was sitting right in front of me and typing away most of the session.