Search International and National Patent Collections

1. (WO2017151466) MODULAR DEEP LEARNING MODEL

Pub. No.:    WO/2017/151466    International Application No.:    PCT/US2017/019599
Publication Date: Sat Sep 09 01:59:59 CEST 2017 International Filing Date: Tue Feb 28 00:59:59 CET 2017
IPC: G10L 15/065
G06N 3/04
G10L 15/16
Applicants: MICROSOFT TECHNOLOGY LICENSING, LLC
Inventors: HUANG, Yan
LIU, Chaojun
KUMAR, Kshitiz
KALGAONKAR, Kaustubh Prakash
GONG, Yifan
Title: MODULAR DEEP LEARNING MODEL
Abstract:
The technology described herein uses a modular model to process speech. A deep learning based acoustic model comprises a stack of different types of neural network layers. The sub-modules of a deep learning based acoustic model can be used to represent distinct non-phonetic acoustic factors, such as accent origins (e.g. native, non-native), speech channels (e.g. mobile, bluetooth, desktop etc.), speech application scenario (e.g. voice search, short message dictation etc.), and speaker variation (e.g. individual speakers or clustered speakers), etc. The technology described herein uses certain sub-modules in a first context and a second group of sub-modules in a second context.