Epistemic markers (epistemic_markers)¶
Language & Meaning module (B6) for hedging / certainty / epistemic marker density.
What it is¶
Deterministic English lexicon + phrase match (T0 / light).
Per-segment marker hits with character spans, speaker/global counts and rates.
Categories:
epistemic_hedge,approximator,modal_uncertainty,certainty_booster.Derived:
hedge_share,booster_share.
What it is not¶
Not
acts.uncertainty(utterance dialogue-act label).Not
ticsfiller/disfluency counts (though seed lists overlap historically).Not ASR confidence /
transcript_qualityword scores.
Language¶
English lexicon v1. Non-English transcripts abstain (usable=false, language_status=unsupported).
Config¶
Owned subtree analysis.epistemic_markers:
min_tokens_for_rates(default 20)enabled_categories(empty = all categories)
Outputs¶
{base}_epistemic_markers.json/.csvSpeaker category-rate charts
Group: additive
epistemic_markers_pooled.by_category+ session bars
Core / extras¶
Core-mode compatible (T0 lexicon, no optional extras, no model downloads). Offline/air-gapped: packaged EN lexicon only.