Open Voice Network Interoperability Project
Incubation
The Open Voice Network Interoperability Project seeks to open and democratize Conversational AI through the definition, development, demonstration, and industry adoption of a universal API for interoperability of natural language/generative AI-enabled conversational assistants. The universal API for interoperability among conversational assistants will enable conversational assistants to marshal the expertise of other assistants in order to assist users in achieving goals that may be outside of the expertise of the first assistant. To do this the API will enable assistants to securely transfer conversations and relevant data to other assistants This ability requires assistants to be able to find each other and to transmit information by means of a common protocol, similar to HTTP for the web. To this end, the Project will develop (but not be limited to) multiple open-source specifications, developer sandboxes, libraries (Python, Java, Node.js, Javascript, Java/Rust, Objective C/Swift, others), and an open-source destination identification and registry system.
This project is lodged at the intersection of natural language AI (processing, understanding, generation) and language model generative AI. According to Precedence Research (Sept., 2023), this "conversational AI" market will grow at a 23.9% CAGR to a total available global market worth $86.4 billion (USD) by 2032. The anticipated specifications for assistant authentication, liveness detection, the protection of user and organizational data, and interoperable system data security speak directly to Trusted & Responsible AI efforts.
The Project emerged from 2017-2018 research by the Massachusetts Institute of Technology (MIT) Auto-ID Center, Capgemini Consulting, and the Intel Corporation. In Q3 2020, it was formally instituted as a project of the Open Voice Network (https://openvoicenetwork.org), a Linux Foundation community founded that year.
In 2023, and as part of the Open Voice Network, the Interoperability Project published and demonstrated the first specifications of "The Message Envelope," which is the working title of the envisioned universal API. As of this writing (October, 2023) the initial specifications are being tested not only by Project participants, but in a proof-of-concept hosted by the Information Systems Authority (RIA) of the Republic of Estonia, which is using the Interoperability Project specifications to exchange text and acoustic dialogs among independent conversational assistants. In addition, a developer sandbox is now being built, and is expected to be active as of January, 2024.
As noted above, the Open Voice Network Interoperability Project seeks to open and democratize the rapidly growing conversational AI industry. As such, it is working with a growing community of participants and developers at the intersection of natural language AI and language model generative AI technology, and addressing trust and ethics issues that are central to the growth and stability of an open, democratized industry, such as assistant authentication, liveness detection, protection of user and organizational data, and interoperable system data security.
The Open Voice Network Interoperability Project sees numerous potential collaboration opportunities with current LF AI & Data hosted projects. These include
- Egeria (location and discovery services)
- Adversarial Robustness Toolkit (security specifications, tools)
- ONNX (best practices for interoperability standards)
- Delta (natural language and speech processing platform)
- The AI Frameworks workstream of the Generative AI Commons
- The Models workstream of the Generative AI Commons
Code is developed under Apache License 2.0 https://github.com/open-voice-network/lib-interop/commit/3b12ea01aa48ed8094b75de5be830a34f82b3d0b
Specifications are developed under the Community Specifications License 4.0 (https://github.com/open-voice-network/docs/blob/5d78ff1ea5d03d94c2f71effe454f2215abfa248/arch-wg/interoperability-community-specifications-license-cs-template.md)
- GitHub (https://github.com/open-voice-network)
- Google Workspace (https://drive.google.com/drive/folders/0AOKRqJFiQYK3Uk9PVA)
- Slack (meeting agendas)
- GitHub (specification development)
- Slack (currently 350 subscribers to Open Voice Network channels)
- Google Workspace (for document, presentation development)
- Zoom
- LF Edge
- LF Hubspot
None.
- Dr. Deborah Dahl, Conversational Technologies Inc., (dahl@conversational-technologies.com), since 2020.
- David Attwater, Talkmap, Inc., (david.attwater@talkmap.com), since 2020.
- Emmett Coin, EJTalk (emmett@ejtalk.com), since 2022.
in development.
Total Number of Contributors to the Project Including Their Affiliations at the Time of Submitting this Proposal
The Project currently lists 43 GitHub contributors.
In development.
https://github.com/open-voice-network/docs/blob/main/arch-wg/code-of-conduct.md
- Maintenance of Open Voice Network Interoperability GitHub repository
- Maintenance of Open Voice Network Google Workspace, for storage of outside research, meeting recordings, and collaborative development of whitepapers.
- Completion and maintenance of the Open Voice Network Interoperability sandbox, now under construction.
https://openvoicenetwork.org/interoperability-initiative/
The Project to date has been governed by the Directed Fund Participation Agreement of the Open Voice Network, last amended 2021.05.13, stored here: https://github.com/open-voice-network/docs/tree/main.
- Schwarz Gruppe, Platinum Sponsor
- Microsoft Corporation, Platinum Sponsor
- Wegmans Food Markets, Gold Sponsor
- Veritone Corporation, Gold Sponsor
- Sestek, Contributor
- Kaizen Secure Voiz, Contributor
- Adobe, Advisor and Participant