Abstract
Long tabletop role-playing campaigns build up hundreds of invented names. General speech recognition mishears exactly these names, so looking something up during play rarely happens. This disclosure describes a read-only companion that listens to a game and shows what the group's own notes say about each name as it is spoken. Speech is turned into text on the player's own device, from the computer's sound (a video call), a microphone (a physical table), or both. Recognition can be primed with the campaign's names. The text is matched against a casebook built from the group's notes database, with entries, aliases and a numbered journal, and refreshed incrementally. Matching uses exact, near-spelling and sound-alike tests, with guards that keep everyday words, real-world places and the players' own characters out. The display shows the names heard, open threads, the matching entry's card, and names heard that the notes don't yet have. Players teach it during play: one selection saves a misheard phrase as an alias for the whole group. After play, the call's transcript drives an AI-drafted update to the notes. That update covers player knowledge only and is written only after a person approves it. The update adds aliases that tune the next session's matching. Audio and live text never leave the device, and the notes are never written to during play. Variations cover other notes sources, audio sources, recognisers and uses, including meetings, events and teaching. The source code is published under the Apache License 2.0.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Bolt, Wayne, "A read-only listening companion that shows a tabletop role-playing group's own campaign notes as names are spoken", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12260