What this category's data looks like
This page addresses the scenario of academic misconduct report investigation and processing. Materials include report materials, investigation reports, and expert appraisal opinions. Sources are multilingual submissions and correspondence, and the update cadence rolls forward as the investigation stages progress. The three types of materials play different roles when entering the document parsing and chunking function: report materials provide original clues and alleged facts, investigation reports carry the investigation process and conclusions, and expert appraisal opinions give professional judgments and their basis. Field characteristics center on names, quantities, dates, and signatures, and the same record must be able to cross-reference each other across different materials. After parsing is complete, parsing and chunking immediately persist to the database as visible data rows and bind a unique data identifier, for use in subsequent retrieval and terminology extraction.
Configuration and implementation checks
This step is executed by parsing and training tasks (server-side tasks), outputting data rows and index status (parsed, indexing, indexed). The unified terminology table and comparison draft are then completed by retrieval and subsequent processing. After parsing is complete, parsing and chunking immediately persist to the database as visible data rows and bind a unique data identifier. The index transitions through the three states parsed, indexing, and indexed. Data pending indexing does not participate in recall. Changing the model only rebuilds data that has already completed indexing.
| Object | Check | Verification action |
|---|---|---|
| Report materials | Fields, dates, and signatures must correspond item by item with the investigation report and expert appraisal opinions, and names and quantities must not appear in two different spellings | Check the file records and data rows separately, and inspect whether the chunks under chunkSize and chunkSplitter retain the date, signature, and receipt fields |
| Investigation report | The name, quantity, and date of the same record are written consistently across the three materials, and the version must be synchronized to the latest | Open the data rows to check the status, and confirm it is in one of parsed, indexing, or indexed and is consistent with the interface display |
| Expert appraisal opinions | Terminology translations can refer back to the original text, the same term is written consistently throughout, and missing signatures or receipts must be marked | Look at the data rows covered by the rebuild task, and confirm that the rebuild corresponding to trainingType only covers data that has already completed indexing |
| Data rows and file records | The identities of file records and data rows are kept separate, and identifiers are not mixed | Check the file records and data rows separately, and confirm that the data rows bind a unique data identifier |
Three common errors
- Phenomenon: The name, quantity, and date of the same record do not match between the report materials and the investigation report. Cause: After chunking multilingual materials, terminology spelling was not unified, and translations cannot refer back to the original text.
- Phenomenon: After the investigation report is updated, the expert appraisal opinions still cite the old version. Cause: The version was not synchronized, and the chunked data rows were not re-persisted to the database when the materials were updated.
- Phenomenon: Materials missing signatures or receipts still enter the index. Cause: Successful parsing is treated as usable, and missing signature and receipt fields were not marked during the chunking stage.
How to verify the setup
Acceptance for this feature: Upload one material, observe the data row status transitions and the coverage scope of the rebuild task, and confirm that the status only changes among parsed, indexing, and indexed, and that changing the model rebuilds only data that has already completed indexing.
Business test: Sample three terms and check whether their spelling is consistent throughout, confirm that translations can refer back to the original text, and record the tokens and array count of the cited text.
- Open the data rows to check the status, confirm it is in one of the three states and is consistent with the interface display, and that data pending indexing does not participate in recall.
- Check the file records and data rows separately, and confirm that the identities of file records and data rows are kept separate, and that data rows bind a unique data identifier.