Controlled Language and Baby Turing Test for General Conversational Intelligence
This work addresses the need for accessible measures and controllable methods in conversational AI, potentially benefiting developers of intelligent assistants, but it appears incremental as it builds upon existing Turing Test and semantic representation concepts.
The authors tackled the problem of measuring and achieving general conversational intelligence by proposing a Baby Turing Test approach and a controlled language based on semantic graphs, aiming to build a general-purpose conversational system like an intelligent assistant for online media and social network data processing.
General conversational intelligence appears to be an important part of artificial general intelligence. Respectively, it requires accessible measures of the intelligence quality and controllable ways of its achievement, ideally - having the linguistic and semantic models represented in a reasonable way. Our work is suggesting to use Baby Turing Test approach to extend the classic Turing Test for conversational intelligence and controlled language based on semantic graph representation extensible for arbitrary subject domain. We describe how the two can be used together to build a general-purpose conversational system such as an intelligent assistant for online media and social network data processing.