AI agent will deliver the best possible experience equal to the one of a human agent
NTT Resonant is using NVIDIA Jarvis, a framework for conversational AI, with NTT Resonant’s conversational AI to build a demo of a freeform restaurant reservation system. The demo provides a full reservation service, from searching for a restaurant to reserving a table, all being done using a freeform spoken conversation with an AI agent.
The demo was initially unveiled during NVDIA GTC21 event held from April 12 to 26, 2021.
NTT Resonant’s investment in AI and R&D has been recognised by international conferences such as the Association for the Advancement of Artificial Intelligence (AAAI) and the International Joint Conference on Artificial Intelligence (IJCAI).
The restaurant reservation system was developed by combining conversational intent comprehension technology ConvBERT and the NVIDIA Jarvis conversational AI framework. ConvBERT is able to comprehend the context of multiple statements with high accuracy and improve comprehension from previous interactions, thanks to NTT Resonant’s proprietary AI algorithms.
In voice conversations with a human, an AI agent responds by applying established rules and using pattern recognition. The demo includes a restaurant search and reservation API that catches necessary information to make a reservation, such as restaurant type and location, time, atmosphere and more, while the human and AI interact freely.
To make the conversation as natural as possible, the AI will insert small talk to create the feeling of actually speaking with a human. NTT Resonant put effort to provide a type of AI that feels human, accurately interprets information and user preferences to provide the best restaurant suggestions, without making the user feel as if they are interacting with a machine.
The demo was developed in English, but NTT Resonant is planning to make it available in Japanese and develop it for other industries beyond restaurant reservation – delivery services, eCommerce, real estate.
The company also aims to work on a platform for multipurpose AI services via APIs, using the ConvBERT API and multiple Resonant AI APIs that will be released.