A Minimal Execution Architecture for Language Interaction
This WRRA Core 2.1 preprint examines offline dialogue control in a Korean café-ordering prototype. It tests which speaker, addressee, and confirmation-question relations are needed to interpret the same spoken assent. It derives fixed-width state encodings of 22 bits for one dialogue and 43 bits for two under the specified task bounds. Offline audio replay and process-termination tests examine recognition deadlines, worker contention, and recovery of persisted approvals without duplicate orders in the tested cases. Validation data and code: https://github.com/Wonsik-Choi-janefather/WRRA-Core-2.1/tree/main/applications/minimal-language-interaction/v0.2.0Earlier study: https://doi.org/10.5281/zenodo.24268233Contact: [email protected]. ORCID: https://orcid.org/0009-0001-4263-9772.
Authors
- Wonsik Choi
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-16
- DOI
- https://doi.org/10.5281/zenodo.24268232
- Primary Topic
- Speech and dialogue systems
- Type
- preprint