STLCA separates what the tenant said, what SDCI later wrote, what the source dataset contains, and what our analysis derives from those records.
DESCRIPTION is treated as the complaint-side narrative. DESCRIPTION1 is treated as the SDCI staff note / handling narrative. We do not merge them into one voice.
A complaint saying “no heat and the elevator has been out for a month” can support two separate issue categories. Issue counts are not complaint counts.
Strict text evidence, borderline matches and City-label inference remain separate. Borderline wording is not promoted into the public strict totals.
Habitability/physical-condition complaints are kept separate from notices, rent, deposits, lease questions, entry disputes and information-only contacts.
A blank public inspection-date field means the field is blank. It does not prove no inspection. A populated date field does not, by itself, prove a physical inspection either.
Rules, accepted and rejected matches, vocabulary variants, borderline cases, source text and QA outputs are retained so a published number can be traced back to the evidence.
The evidence supports: 9,074 closed/completed strict habitability records have no inspection date in the City's public-data field.
Closed/Completed is a case status. Correction and compliance verification require separate evidence.
The analysis expressly separates paperwork, inquiries, staff notes and physical-condition complaints.
Future RRIO/report comparisons will classify supported, discrepant, not-comparable, incomplete and unresolved results. Intent requires separate evidence.
| Layer | What it contains | How it is used |
|---|---|---|
| City public export | Record identifiers, type, status, dates, address, complaint/staff text and public inspection fields | Source record and join backbone |
| Complaint parser | Issue categories, authorship/channel signals, evidence strength, flags and vocabulary matches | Derived classification; never replaces source text |
| Portal reconstruction | Timeline, related records, inspection details, findings documents and enforcement evidence where extracted | Used to test what a high-level public field actually represents |
| RRIO / permits / property joins | Property-level regulatory and permit chronology | Pending QA Not yet used for the public findings on this release |