Output database
The normal run writes:| Table | What it stores |
|---|---|
run_metadata | One row per run_id: wrapper commit, dirty flag, cTAKES version, dictionary hash, pipeline hash, Java version, command line, input/output folders, writer profile, manifest path. |
documents | One row per note: document_id, text length, pipeline name, processed time. |
concepts | Clinical concept rows with span, text, section, CUI, RxCUI, preferred text, TUI, semantic group/type, assertion fields, POS, relation fields, and WSD fields. |
rxnorm | Medication mention rows with span, CUI, RxCUI, RxNorm name, dose, route, frequency, form, duration, status change, allergy, start date, and end date. |
cui_counts | Per-document CUI counts split into affirmed and negated. |
CSV/XMI output
With--writers standard, the output folders are:
| Folder | Writer |
|---|---|
xmi/ | FileTreeXmiWriter. |
concepts/ | ConceptCsvWriter. |
rxnorm/ | DrugRxNormCsvWriter. |
cui_counts/ | CuiCountSummaryWriter. |
DrugRxNormCsvWriter writes one *_drug.csv per note. The medication columns are:
Run manifest
Every real run writes:| Field group | What it proves |
|---|---|
run_id, started_at, status | Which execution this was and whether it completed. |
documents | How many notes were planned and completed. |
pipeline | Pipeline key, source .piper, effective .piper, and hashes. |
wrapper | Repo path, git commit, and dirty flag. |
ctakes | cTAKES version and runtime home. |
dictionary | Source XML, effective XML, XML hashes, DB hash, and isolation reason. |
java | Java binary and Java version line. |
input | Requested input folder and runtime/sanitized input folder. |
output | Output folder, work folder, writer profile, and SQLite path. |
invocation | Sanitized command args. |
run_manifest.json, run_pipeline.effective.piper, and run_dictionary.effective.xml.