Irony and Sarcasm
What it is: Saying the opposite of what is meant, often with a tonal cue.
Example: “Oh, fantastic job…” said with clear frustration.
Why machines miss it: Literal interpretation of words leads to mislabeling intent.
Pragmatic Implicature
What it is: Inferring meaning beyond explicit words, based on context.
Example: “It’s cold in here” might mean “please close the window.”
Why machines miss it: Requires theory of mind — understanding what the speaker intends, not just what they say.
Prosody and Micro‑Pauses
What it is: Subtle pitch, rhythm, and pause patterns that change meaning.
Example: A pause before “right…” signals doubt; no pause signals agreement.
Why machines miss it: Text transcription drops acoustic cues unless explicitly annotated.
Politeness Strategies
Example: “Would you mind…” vs. “Do this now.”
What it is: Choosing indirect or softened language to show respect or reduce friction.
Why machines miss it: Cultural norms for politeness vary, and LLMs lack lived experience to interpret subtleties.
Conversational Turn‑Taking
What it is: Signals (intonation, timing, filler words) that indicate when one speaker yields or holds the floor.
Example: “Uh — well, I think…” might indicate hesitation.
Why machines miss it: Requires awareness of discourse patterns, not just sentence-level analysis.
Emotion Through Nonliteral Cues
What it is: Expressing feelings indirectly via metaphor, humor, or understatement.
Example: “I’m just over the moon” meaning deeply happy.
Why machines miss it: Idiomatic language often bypasses literal models.
Cultural Variation in Directness
What it is: Some cultures value blunt clarity; others favor indirect phrasing.
Example: In some contexts, “That’s difficult” really means “No.”
Why machines miss it: Requires sociolinguistic knowledge and cultural context.
Focus and Emphasis
What it is: Shifting meaning with stress or tone within the same sentence.
Examples:
- “I didn’t say he stole the money,” meaning you might have implied it, but you didn’t state it specifically.
- “I didn’t say he stole the money,” meaning that the money was stolen, but you didn’t accuse him.
- “I didn’t say he stole the money,” meaning he might have borrowed it or found it, an alternative to theft.
Why machines miss it: Text alone lacks the phonetic cues that disambiguate emphasis.
Shared Knowledge and Presupposition
What it is: Assuming the listener already knows certain facts.
Example: “You remember the meeting…” implies context not stated.
Why machines miss it: Requires integrating prior conversational, relational, or organizational context.
Indirect Refusal or Agreement
What it is: A response that sounds positive but implies rejection.
Example: “I’ll think about it…” often means “no.”
Why machines miss it: Needs pragmatic interpretation and awareness of subtle conversational norms.
If you’re looking to ensure the full meaning of language is transcribed, talk to an expert at Sigma to learn how our annotators can help.