OneConnect US Research Institute (GammaLab NYC), has placed first in the seventh Dialog System Technology Challenge (DSTC7 or “the Challenge”), one of the world’s leading dialog system challenges. GammaLab NYC achieved the highest human rating score on appropriateness and the second-highest human rating score on informativeness with its OneConn-MemNN System for Knowledge-Grounded Conversation Modeling.
A leading academic competition in the dialogue system of AI, the Challenge was started by scientists from Microsoft, Carnegie Mellon University, and the Honda Research Institute. This year, the Challenge was split into three different tracks, with GammaLab NYC participating in the Sentence Generation track, where performance was evaluated based on appropriateness, informativeness and utility.
Read More: Why Facial Recognition Providers Must Take Consumer Privacy Seriously
For dialogue systems such as chat bots, a main challenge for developers remains how to combine the dialogue content with the exact context to avoid generating unclear and irrelevant sentences. The “OneConn-MemNN System” has an advantage in overcoming this challenge. The model integrates a large number of external knowledge and human corpora and builds end-to-end neural approaches for conversational response generation based on deep learning. It can generate smooth, coherent, interesting, and informative responses. Meanwhile, the system uses a mechanism to process key information in the conversation to generate different responses.
Read More: Jumpstart 2019: Interview With Tracy Malingo, SVP of Product Strategy at Verint
This technology could be widely used in various kinds of business scenarios that can leverage AI dialogue technology and natural language generation (NLG). For example, in a smart customer service system, it can generate coherent and interesting conversations based on the real-world knowledge it picked up to improve user experience. It can also help automatically transform massive structural data in fields such as insurance , banking, investment, and medical care, into human sounding narratives with different writing styles to greatly reduce manual reporting efforts and improve decision making.
GammaLab NYC will present the OneConn-MemNN System in the DSTC7 workshop of the 2019 Association for the Advancement of Artificial Intelligence Conference (AAAI-19), one of the top AI conferences in the world in January 2019.
Read More: Interview with Cédric Carbone, CTO at Ogury
Scrap copper communication Scrap Copper recycling machines Metal waste reprocessing center
Scrap Copper cable sorting, Metal recycling and reprocessing, Copper scrap alloying