WorldsBestChatbot.com
A new frontier in Artifical Intelligence
  Print Version 
 

Flirting Chatbots

In December 2014 New York Magazine published an article by Melissa Dahl with the provocative title "Computers are Better at Flirting than We Are. Dahl’s article summarizes some work by three Stanford University researchers in which they developed a computer program that is “pretty good” at picking up on flirtatious behaviour by humans. One of the Stanford trio, Dan Jurafsky, designed a flirtation-detection system based on established machine learning techniques. The learning system was trained on the transcripts of 1,100 speed dates, each lasting 4 minutes.

The Stanford researchers concluded their detailed report on their work1 by describing their system as being
". . . able to predict flirtation intention better than humans can, despite humans having access to vastly richer information (visual features, gesture, etc.). This system facilitates the analysis of human perception and human interaction and provides a framework for understanding why humans perform so poorly on intention prediction


In order for a computer program to be able to flirt successfully with a human conversation partner it is not only necessary for the software to be able to detect flirtatious behaviour by the human. It must also be able to engage in flirtatious conversation itself. We plan to develop two flirtatious versions of our chatbot, one with a male persona and the other a female persona. At the same time we will be putting Do-Much-More through an extensive overhaul, with a regime of upgrades that will result in a quantum leap in conversational performance.

For more news about these exciting developments please visit us again from time to time.


1"It's Not You, It's Me: Detecting Flirting and its Misperception in Speed-Dates." Rajesh Ranganath, Dan Jurafsky & Dan McFarland. Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1, pp. 334-342.


All text and images on this site are copyrighted and cannot be reproduced without permission.
Copyright © 2017 David N. L. Levy, 2017, built by Scanraid Ltd