Tencent Gander is an AI assistant that works and speaks at the same time.

Gander is a multi-functional task-oriented dialog system with offline capability. Gander is composed of “cerebellum” and “brain”, where “cerebellum” handles one second of speech at a time, and “brain” is responsible for handling complicated tasks in a modularized way. At present, our model doesn’t perform best (a little worse than other models), which we think is caused by the error in speech recognition. We will open source our code on github now, and the weight file and data will be published later.

Compared with current voice assistants that interrupt users when they receive new requests, Gander seeks to solve the problem.