A water utility wants work-note extraction that understands meter-related jargon such as register freeze and AMI hop failures. What customization should they apply?

The water utility wants work-note extraction that understands meter-related jargon such as register freeze and AMI hop failures. They should customize domain extraction so operational work notes map those terms into structured fields rather than treating them as noise. Domain customization—glossaries, few-shot examples, or fine-tuned extractors—teaches the model the utility’s shop dialect so maintenance systems receive usable entities instead of mangled free text.

Customizing intent detection for utility meter jargon in operational work notes uses the right vocabulary on the right notes but the wrong task, since intent classification labels the request purpose rather than extracting terms like register freeze or AMI hop failure into structured fields. Customizing domain extraction for utility meter jargon in engineering design specs keeps meter vocabulary but targets the wrong source, so its model expects design-spec language while meter terms like register freeze and AMI hop failures in operational work notes remain unmapped. Customizing domain extraction for utility meter jargon in annual customer reports keeps meter vocabulary but targets the wrong source, so its model expects summarized customer-facing language while meter terms like register freeze and AMI hop failures in operational work notes remain unmapped. Customizing domain extraction for utility meter jargon in annual customer reports fails because the stem asks for extraction from operational work notes, not from customer-facing summaries.

Exam caveat: intent detection still misses extraction needs; when the stem highlights meter-jargon extraction from work notes, the answer is meter-domain customization of extraction in operational work notes, not intent detection and not extraction tuned for engineering design specs or annual customer reports. Also distinguish meter extraction in operational work notes from intent detection and from meter extraction in design specs or customer reports.

Operational check: build a labeled set of meter work notes containing register freeze, AMI hop failures, and related codes; evaluate precision/recall per field; involve field supervisors in error review; and promote the customized extractor only after structured outputs land correctly in the work-order system. Monitor drift as new meter firmware introduces new failure phrases.