Project Expo
Project Profile
Project abbreviation: COMPRISE
Project name: Cost-effective, Multilingual, Privacy-driven voice-enabled Services
Project coordinator: Emmanuel Vincent (emmanuel.vincent@inria.fr)
Project consortium:
- Inria
- Ascora
- Netfective Technology
- Rooter Analysis
- Tilde
- Universität des Saarlandes
Funding: 3.2 Million €
Project duration: 12/2018 – 11/2021
Main key words: Voice Assistants, ASR, SLU, MT, Privacy, Multilingualism, Platforms
Background of the research topic: While voice-enabled applications are becoming ubiquitous, their development requires increasing amounts of human-labeled voice and text data and must be repeated in every language. This raises critical privacy concerns, incurs high costs for voice technology companies and application developers, and fails to address all languages and users.
Goal of the project: COMPRISE has contributed to building voice interaction technologies that are more secure, more cost-effective, and more inclusive.
Project abstract: To do so, COMPRISE has released a series of innovative software tools to hide personal attributes in voice and text data, train automatic speech recognition and natural language understanding systems from partially unlabeled data, and quickly expand their usage to multiple languages or accents. These tools have been integrated in the COMPRISE SDK and the COMPRISE Cloud Platform, which provide full-fledged functionality for voice technology companies and mobile application developers.
Publications: https://www.compriseh2020.eu/