Predicting detection filters for small footprint open-vocabulary keyword spotting

In many scenarios, detecting keywords from natural language queries is sufficient to understand the intent of the user. In this paper, we propose a fully-neural approach to open-vocabulary keyword spotting, allowing a user to include a voice interface to its device without having to retrain a model on task-specific data. We present a keyword detection neural network weighing less than 550KB, in which the topmost layer performing keyword detection is predicted by an auxiliary network, that may be run offline to generate a detector for any keyword.