Academica keeps every original source file unchanged in an archive. A source only enters the searchable database after it is parsed into clean records. The shape of those records decides exactly where they go.
The record's shape decides its destination
downloaded source → a document: shared library of documents → a record matching an existing shape: existing table → a record with a new shape: its own table → not yet parsed: stays in the archive
One job for each destination
| Destination | Rule | Examples |
|---|---|---|
| Shared document library | One entry per document version | Guidance, assessments, reports |
| Existing table | The record fits a shape already tracked | Trials, safety reports, prices, awards |
| New table | No current shape preserves its meaning | Distinct structured feeds |
| Archive only | Raw file with no parsed records yet | Unparsed releases |
Parsing comes before searchable storage
- A file that cannot be read into records is not added to the database.
- The archive, not the database, holds the raw original bytes.
- A new table is created only when existing shapes would lose meaning.
- Every parsed record keeps a pointer back to its archived original.
Families of evidence share the same rules
| Evidence family | Example sources | Destination |
|---|---|---|
| Literature and citations | Europe PMC, Semantic Scholar, OpenAIRE, CORE, preprints | Articles table or shared document library |
| Trials and grants | CTIS, ISRCTN, ICTRP, NIH RePORTER | Trials or grants table |
| Guidance and assessment | NICE, USPSTF, FDA committees, health-technology agencies | Shared document library |
| Prices and payment | Transparency files, hospital prices, national fee schedules | Pricing and coverage tables |
| Statistics and classifications | OECD, Eurostat, Orphanet, national statistics | Statistics or rare-disease tables |
| Procurement | UK notices, China procurement, KHIDI | Procurement table |
| Device authorization | PMDA, NMPA, MFDS, HSA | Device authorizations table |
Regional sources reuse common shapes
Japan, China, Korea, Singapore, and Australia release different files but the same recurring record forms. Device authorizations settle into one shared shape. Coverage and fee rules use one shared payment shape. Narrative guidance stays in the document library.
126 source families tracked
4 main destinations
1 archive of originals
When a new table is worth it
- Reuse an existing table when its key and time period match the source.
- Use the shared document library for narrative publications.
- Create a new table only for a distinct, reusable record shape.
- Keep unparsed families in the archive until they can be read into records.