Mentions & Social · Signal-Driven#280

Scripted-vs-Q&A Segment Split

The same speaker uses very different vocabulary in a scripted segment than in unscripted question time — earnings calls are the clearest case, where the prepared statement is lawyered and the analyst Q&A is loose. A word that will never appear in the script frequently appears in the answers. You model the two segments separately, and during a live event you reprice sharply at the moment the format switches.

What you need to run it

Segment-tagged transcript corpus separating prepared remarks from open questions
Per-segment word frequency rates for each speaker
Live detection of the transition point during the event itself

Where this applies

Markets on Polymarket where scripted-vs-q&a segment split is the natural play:

Will the CEO mention layoffs on the earnings call?
Will the official say 'recession' during the press conference?
Will the term come up during the question period?

Capabilities this demands

Data ingestionFeed ingestionModel / quant

At a glance

CategorySignal-Driven

MarketMentions & Social

Requirements3

CapabilitiesData ingestion, Feed ingestion, Model / quant

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 feeds directory.