In an effort to enhance the use of artificial intelligence (AI) in the defense sector, the European Commission (EC) has chosen a number of collaborative AI research and development projects under the European Defense Fund (EDF) 2021 Calls for Proposals.
After an open call for innovative defense solutions by small and medium-sized enterprises (SMEs), the EC selected ALADAN (“Ai-based LAnguage technology development framework for defense ApplicatioNs”), among other projects.
The project started in December 2022 and will run for 42 months until 2026. Total funding amounts to EUR 3.2m.
The EC’s project fact sheet said that, while AI language technology has had considerable success in “civilian” applications, its development in defense has been impeded by a number of barriers; namely, data confidentiality, and sector-specific terminology.
Training AI models requires a huge annotated corpus. Since such data is generally classified, its preparation and use are subject to significant constraints. As a consequence, the development of AI models can be expensive and slow. Moreover, the resulting models are classified and cannot be mutualized.
ALADAN intends to design and develop a disruptive framework for developing AI-based language solutions for defense applications. Four language technologies will be addressed: spoken language identification, speech recognition, spoken term search, and text and speech translation.
The novel framework will rely mainly on open-source data from multiple domains, while targeting in-domain data, using only small amounts of application-specific data for validation. According to the project fact sheet, the project will enable the development of an in-domain solution within 6 months.
The project has several intended outcomes. Firstly, the initiative will lead to language processing solutions that cover multiple domains. Secondly, the project will validate the new development framework. Thirdly, an easy-to-use toolkit will be created that enables non-expert users to customize models using classified data.
A further outcome, according to the EC, will be the development of a demonstrator that combines the targeted technologies with multi-domain models for at least two defense use cases.
According to the fact sheet, the resulting open-domain development framework and adaptation toolkit will be “game changers in the adoption of language technologies by the forces.”
Furthermore, ALADAN will “reduce the costs (time and financial) of developing language technologies for military applications, thereby drastically reducing the barriers for their wider uptake,” according to Vocapia Research.