Comment by charliereese

We process podcast transcripts and other datasets, use that to match over time based on similarity in meaning (or optionally exact and fuzzy matches if desired), then feed matching excerpts (match + surrounding context) into an LLM to score whether the speaker / writer agrees with {customizeable_poll_query_and_scoring}

:)

Note: similar to how DeepSeek started (or pandas or pick one of the many data science innovations), we actually are a quant investing software and data company. As a result of that, everything is time-series.