liveBook · Manning

Unintelligible utterance

As explained in Effective Conversational AI

An unintelligible utterance is conversational input whose meaning cannot be reliably determined because it is incoherent, severely corrupted, or otherwise incomprehensible.

An unintelligible utterance is input whose meaning cannot be determined reliably. In conversational logs, it may result from keyboard mashing, severe typos, speech-to-text errors, or captured background conversation. section 4.4.2

Recognizing the failure

Inputs such as “does it school,” a long fragment from a caller’s background conversation, or “klewtkhaccalifornia liense” are too incoherent for dependable intent assignment. section 4.4.2

Handle suspected unintelligible utterances during candidate review:

  • Make a quick usability decision rather than over-analyzing; discard the input if no determination is possible within about a minute. section 4.4.2
  • Exclude clearly unintelligible inputs from blind test sets. section 4.4.2
  • Set recurring speech-transcription or environmental-noise patterns aside for further evaluation or referral to the appropriate team. section 4.4.2

How Effective Conversational AI covers it

How deep it goes
Defines causes, gives three examples, and specifies rapid review, exclusion, and escalation steps.
Examples
Uses incoherent speech, captured background conversation, and severe typos as representative examples.
What you can do after
Diagnose unusable log inputs, exclude them from blind tests, and escalate recurring transcription or noise patterns.
How it connects
Connects unintelligible utterances with speech-to-text, background noise, production logs, and blind testing.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Effective Conversational AI. The section links point to the parts of the book it draws on.