Natural Language Dialogue

- Toward the Goal of Passing the Turing Test

Natural language dialogue between human and computer is regarded as one of the most challenging problems in artificial intelligence. The Turing test precisely takes this task as a benchmark to verify a machine's ability to exhibit intelligent behaviors equivalent to or indistinguishable from that of a human. Researchers in the lab are also tackling this grand challenge, specifically they are trying to study the feasibility of building a system by using a vast amount of data and machine learning techniques.

Multi-turn dialogue is certainly very difficult, and thus the current focus of research in the communities is single-turn dialogue. The researchers in the lab have, perhaps for the first time, conducted a comprehensive investigation on building a single-turn dialogue system, with deep learning techniques. Several models based on deep learning and big data have been developed, which can make the system return reasonable output to human’s input in more than 70% of the cases. The interesting question, then, is: what performance can the system achieve with more advanced deep learning techniques and much more data?

A task on single-turn natural language dialogue using conversational data available on social media, referred to as short text conversation, is being organized at the NTCIR conference. We wish to work jointly with the community to take one more step toward realizing this very dream of AI.