Modular deep learning model

US10235994B2 · kind B2 · utility

Assignee

• MICROSOFT TECHNOLOGY LICENSING, LLC · US

Inventors

• Yan Huang · Hsinchu, TW
• Chaojun Liu · Redmond, US
• Kshitiz Kumar · Redmond, US
• Kaustubh Kalgaonkar · Atlanta, US
• Yifan Gong · Sammamish, US

Key dates

• Filing date: Jun 30, 2016
• Grant date: Mar 19, 2019
• Expiry date: Jun 30, 2036

Classification

• Technology area (CPC G): Physics
• CPC primary: G10L15/28
• WIPO field: Computer technology
• WIPO sector: Electrical engineering

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.