NLP (natural language processing) - has anyone done this?

Solar2D General Questions/Discussion
#1

Hey,

Im looking for the option to parse an input text for a yes/no/maybe questions, in a way that the user 

enters free text as an answer to a question, and the algorithm can understand if he’s meaning was “yes” or “no” or “maybe” or “not understood”. 

Do you guys know of a good way to start? i remember even 15 years ago there were chat bots that could understand your answer to a yes/no question.

I wonder if anyone has done this kind of thing in LUA… or is there an open source library for this kind of thing, I’ve looked at Stanford CoreNLP but it seems way too much advanced to my requirements.

Any help would be greatly appreciated.

Roy.

#2

I did something like this on the zx spectrum when I was a kid to simulate A.I.

What I did was create a table of all the possible answers a question would provoke then looking for certain keywords by scanning the line of text the user inputs. I would narrow a user input string down to two or three key words that I was expecting to see to a given question. So the order in which they typed their answer didn’t matter.

This is easily done with the string API in Corona here: https://docs.coronalabs.com/daily/api/library/string/index.html

Hope this gives you ideas.

#3

@QuizMaster thank you for the suggestion.

Yeah, a table with optional answers was my direction.

Roy.

#4

I did something like this on the zx spectrum when I was a kid to simulate A.I.

What I did was create a table of all the possible answers a question would provoke then looking for certain keywords by scanning the line of text the user inputs. I would narrow a user input string down to two or three key words that I was expecting to see to a given question. So the order in which they typed their answer didn’t matter.

This is easily done with the string API in Corona here: https://docs.coronalabs.com/daily/api/library/string/index.html

Hope this gives you ideas.

#5

@QuizMaster thank you for the suggestion.

Yeah, a table with optional answers was my direction.

Roy.