3
$\begingroup$

I am building an artificial town that has randomly generated people with different personalities. People in this artificial town have attributes that change their happiness, ego, rep, etc. I already have most of the code done, but I ran into a problem. I don't know how I could generate conversations between two people.

Does anyone know a specific AI/NLG technique that I could use to accomplish that?

My knowledge in NLG used to be pretty vast, but I haven't worked with NLG or NLP in over a year, so I'm a little rusty. The only thing I need to make this a success is for the people to generate their own dialogue based on where they are or who they are talking to.

Some examples of dialogue that I would want to be generated are:

Peter: Hello, John. How is your wife Jane?

John: Hello, Peter. My wife Jane is doing great!

Or

Peter: What's your child's name?

John: My child's name is Johnny.

The agents will have access to the datasheets with all of the names, emotions, and reputations on it. I'm hoping to build a datasheet for each person, so that they only know as much data as they learn from the dialogue they have with the other people who know that data. The dialogue should not repeat as each dialogue should act as a sort of random get request.

$\endgroup$
4
  • $\begingroup$ This is a really copmplex, difficult problem in general. If you could give more information about your setup, your existing knowledge of natural language generation, and what sort of thing would count as successful, that may help. At least it may make it possible to answer the question. However, be prepared to adjust your expectations . . . $\endgroup$ Oct 27 '20 at 18:48
  • $\begingroup$ For instance, you could just find or build a big database of conversations and sample from it, if the point is to have the characters "looking active". Whilst if you want them to talk about something relevant to their current situation (including references to objects or other people in the town) or be original and entertaining, then that may be beyond current cutting edge $\endgroup$ Oct 27 '20 at 18:51
  • $\begingroup$ Facebook made a negotiating bot that used words. towardsdatascience.com/… Funny question is why fb wants to simulate a human conversation that wasn't between humans. I think they fake some of the interactions using bots meant to mimic your friends in order to turn FB into a more addictive slot machine. $\endgroup$ Oct 27 '20 at 18:56
  • $\begingroup$ @NeilSlater I have it setup as objects whole have attributes that change their happiness, ego, rep, etc. My knowledge in NLG used to be pretty vast but I haven't worked with NLG or NLP in over a year so I'm a little rusty. The only thing I need to make this a success is for the people to generate their own dialogue based on where they are or who they are talking too. $\endgroup$ Oct 27 '20 at 22:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.