Researchers at the Osaka University, Japan, have designed a methods for conversational systems – the highly advanced computer systems that are intended to talk with humans in natural languages. The new method allows the system to learn or understand the ontological category of an unknown term in a conversation with the use of implicit confirmation instead of responses to explicit questions that could turn out to be very abrupt.
A number of chatbots, conversation robots, and voice assistant applications have been introduced in the past few year. However, these systems essentially answer queries on the basis of preprogrammed data. Another method through which computers learn from humans is by posing simple questions in a repetitive manner. However, when the questions are abrupt or the same over a long duration, users are likely to lose interest in the conversation with the system.
The group of researchers from Osaka University have developed a method in which the computer systems confirm whether its understanding of an unknown word is correct by making confirmation request to the user. When the user responds to these requests, the system gathers the required knowledge regarding new words during an interaction. Using the response of the user that follows after every request and the context of the conversation, the system makes a decision whether its prediction has been correct or not.
Chatbots in the market presently converse with anyone in the same way. But as dialogue systems gain increasing popularity among users, computers will have to converse by learning with a user according to the context and situation. The results from this study could prove to be a new approach in the direction of dialogue systems that can make computers smarter through its conversations with humans and can lead to the development of a smart dialogues system that has the ability of customizing responses according to user’s situation.