NLU Framework for Voice Enabling Non-Native Applications on Smart Devices

摘要

Voice is a critical user interface on smart devices (wearables, phones, speakers, televisions) to access applications (or services) available on them. Unfortunately, only a few native applications (provided by the OS developer) are typically voice enabled in devices of today. Since, the utility of a smart device is determined more by the strength of external applications developed for the device, voice enabling non-native applications in a scalable, seamless manner within the device is a critical use case and is the focus of our work. We have developed a Natural Language Understanding (NLU) framework that uses templates supported by the application (as determined by the application developer). This framework can be employed in any mobile OS (Android, iOS, Tizen, Android wear) for a wide range of devices. To aid this demonstration, we have implemented the framework as a service in Android OS. When the user issues a voice command, the natural language query is obtained by this service (using one of local, cloud based or hybrid speech recognizers). The service then executes our NLU framework to identify the relevant application and particular action details. In this demonstration, we will showcase this NLU framework implemented as an Android service on a set of applications that will be installed on the fly. Specifically, we will show how the voice queries are understood and necessary services are launched on android smart wearables and phones.