Sorry for the flood of posts, I’m running through a list I’ve been keeping.

Question du moment regards garbage input.

One issue I’ve been having is that players testing my bot will occasionally type total nonsense just to see the response. When this happens in a gambit (and my bot is basically a series of forked gambits), the next response fires a little too naturally/easily and I get exchanges like:

Renata1: Follow my directions exactly.
rob:_> lksdgfkdshfd
Renata1: Go to https://www.facebook.com/
rob:_> sdlkjfhsdkfhsflkh

Is there any way to identify single-word inputs that aren’t in wordnet or any of the substitution dictionaries as “garbage”? I assume this would have to first be done on the concept level and then I could form a topic around that to catch this kind of input and respond accordingly. The trouble is that I’m unclear on how to frame such a concept.

Anyone have any experience or thoughts here?

Thank you as always,
Rob

The answer to a bunch of posts is “learn to use :prepare”.  This will tell you how the system interprets user input (without reacting to it).  It would tell you that Lydia Linder gets name merged into a single token. It will tell you which form a word (lower or upper case) is chosen for original and canonical.  And it will tell you what concepts it marks for each word.

In the case of junk words like above…. ~unknownword is likely to get flagged.

Duly noted. Thanks again.

I categorise any word of 8 or more letters with no vowels as junk which manages to stop of lot of garbage. Not sure if you can do this in Chatscript.

“You can do anything in ChatScript if you want to write enough script”.

Chatscript labels all words not found in the dictionary as ~unknownword.  You can convert any such matched word into its set of letters (as facts) and analyze the letters in any way you choose to decide it is garbage.