Hokkien Low Density Language Capability

Black Cape plans to develop an application that performs speech-to-speech transcription and collects and preserves spoken text that can be leveraged for follow-on model improvements. Encapsulating the majority of overall Black Cape system complexity within a single semantically versioned Docker image will allow for continuous improvement of a low-density language translation container while keeping the service-to-service level API used by USSOCOM’s field platform services uniform and stable. Once a stable API layer is in place, all other components of the system (e.g., base translation model, any possible speech-to-text models, voice encoder models, required runtime libraries, threading, or worker scaling solutions) can be iteratively improved after initial fielding without disrupting the overall function of the USSOCOM translation platform.