What this category's data looks like
This page addresses the material verification and discrepancy localization scenario for peer review organizations, covering three types of materials: review invitations, expert review comments, and author revision notes. Review invitations record the review target, manuscript number, invitation date, and receipt status; expert review comments carry the review conclusion, comment body, submission date, and signature; author revision notes correspond to point-by-point responses, revised manuscript version, date, and receipt. The material sources are editorial office correspondence emails, review system exports, and author submissions; the update cadence advances with review rounds, and versions iterate frequently. The three types of materials play different roles when entering the document parsing and chunking function: review invitations and revision notes mostly serve as comparison baselines, while expert comments serve both as a baseline and as an object to be verified; after chunking, each forms retrievable data rows for subsequent retrieval and discrepancy localization calls.
Configuration and implementation checks
This step is executed by the parsing and training task (server-side task), outputting data rows and index status (parsed, indexing, indexed). The "discrepancy list" is 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 three states, and data pending indexing does not participate in recall.
| Object | Check | Verification action |
|---|---|---|
| Review invitation | Whether the manuscript number, review date, and receipt status of the same manuscript are consistent across the two invitations | Check the file record and the data row separately, inspect the chunks corresponding to trainingType and chunkSize, and confirm the status is parsed or indexing or indexed |
| Expert review comments | Whether the review conclusion, number of comments, submission date, and signature in the two comments match up | Compare the two materials field by field, check whether the chunkSplitter split boundary falls between comment entries, and record inconsistent fields |
| Author revision notes | Whether the version number, number of responses, date, and receipt of the revision notes and the revised manuscript are in sync | Check cases where the version is not in sync and where the signature or receipt is missing, and confirm that the rebuild task only covers data rows whose indexing is completed |
| Discrepancy list | Whether the list fields correspond one-to-one with the original fields of the two materials with no omissions | Hand over to the reviewer for confirmation and write back; before writing back, check that the data row status is consistent with the interface display |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match across the two materials. Cause: The material versions are not in sync; the old and new versions were parsed and persisted at the same time, and each became its own row after chunking.
- Symptom: The review comments lack a signature or receipt, yet the discrepancy list shows them as complete. Cause: The signature page became its own chunk and was not associated, so retrieval did not cover that chunk.
- Symptom: A just-uploaded material cannot be found. Cause: The data row is in indexing status, and data pending indexing is naturally isolated at the retrieval layer and is not recalled.
How to verify the setup
This feature's acceptance: upload one material, observe the data row status transitions and the coverage scope of the rebuild task, and confirm the status is parsed, indexing, indexed in sequence, and that the rebuild targets only data whose indexing is completed.
Business test: use one real record to find fields inconsistent with the ledger, and check the name, quantity, date, signature, and receipt item by item.
- Open the data row to check the status, check the file record and the data row separately, and confirm that the identifiers of the file and the data row are distinguishable
- View the data rows covered by the rebuild task, confirm that only data whose indexing is completed is covered, and record the actual values of
chunkSize,chunkSplitter, andtrainingType