Mentions & Social · Signal-Driven#224

Live-Transcript Word-Count Race

Mention markets ask whether a specific person says a specific word during a speech, hearing or broadcast, and they resolve on the transcript. Automated speech-to-text on the live audio stream gives you the count seconds before human viewers register it and well before the official transcript posts. You run recognition on the feed, match against the market's exact target phrase, and fire immediately on a confirmed hit.

What you need to run it

Real-time speech-to-text on the live audio or video stream
Exact phrase-matching that respects the market's stated definition, including variants
Confidence thresholds so a mis-transcription doesn't trigger a bad fill

Where this applies

Markets on Polymarket where live-transcript word-count race is the natural play:

Will the president say 'inflation' during the State of the Union?
Will Elon Musk mention Mars on the next earnings call?
Will the Fed chair say 'recession' in the press conference?

Capabilities this demands

Low latencyFeed ingestionCustom code / API

At a glance

CategorySignal-Driven

MarketMentions & Social

Requirements3

CapabilitiesLow latency, Feed ingestion, Custom code / API

VenuePolymarket (CLOB, Polygon)

Build it

Related mentions & social strategies

This is documentation, not advice. Poly Research & Robotics publishes how these strategies work because the method should be checkable — not as a recommendation to trade them. See the full strategy database (297 strategies) or the data resources directory.