Knowledge Base Retrieval for Peer Review Organization: Compare Documents and Locate Differences

Peer Review: review retrieval for document checks. Compare the source records, required inputs and downstream results before handing the task over.

What this category's data looks like

This page addresses the material verification and discrepancy location scenario for peer review organizations. Materials come from three types of documents: review invitations, expert review comments, and author revision notes, and enter the knowledge base in a versioned manner, with update cadence synchronized with manuscript flow. Field characteristics include name, quantity, date, and signature fields; the name, quantity, and date of the same record may not match across two materials, versions may be out of sync, or signatures and receipts may be missing. When entering the knowledge base retrieval and recall function, review invitations serve as object identifiers and time anchors, expert review comments serve as judgment basis and opinion fields, and author revision notes serve as receipts and revision items. The three types of materials are recalled as candidate snippets, similarity scores are used for ranking, and the final discrepancy list is generated by subsequent nodes or manually based on this.

Configuration and implementation checks

This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate snippets and similarity scores (after pattern, merge, rerank, filter, and limit trimming). The discrepancy list is then completed by subsequent nodes or manually.

ObjectCheckVerification action
Review invitationWhether the name, quantity, date, and signature fields are consistent across the two materialsIn the search test, test once each with synonymous questions, switch searchMode to semantic, full-text, and hybrid, compare the number of candidates and scores, and check the name, quantity, date, and signature fields in the invitation letter
Expert review commentsWhether the name, quantity, and date of the same record are consistent with the ledger, and whether the version is synchronizedTurn usingReRank on and off to compare the ranking of the top few items, check the date and version number of the opinion fields, and record the similarity score and the number of candidates
Author revision notesWhether the number of revision items, dates, and receipt signatures correspond to the review commentsAdjust limit to observe changes in the number of candidates, check the quantity, date, and signature receipt in the revision notes, and separately record the cited text tokens and array item count
Discrepancy listWhether field-level discrepancies can be located to specific materials and rowsCompare the candidate snippets with the ledger field by field, mark inconsistent items for name, quantity, date, and signature, and submit them to the reviewer for confirmation and write-back

Three common errors

  • Symptom: The name, quantity, and date of the same record do not match across two materials. Cause: Candidate snippets come from different versions and were not aligned and checked by field.
  • Symptom: Versions are out of sync, and the dates in the revision notes and review comments contradict each other. Cause: Versions were not distinguished during recall, and field-level comparison was not performed after candidate merging.
  • Symptom: Signatures or receipts are missing, and the discrepancy list cannot be closed. Cause: Signature fields were missing when the materials entered retrieval, and candidate snippets did not cover that field.

How to verify the setup

This function's acceptance: test once each with synonymously rewritten questions, compare candidates and similarity scores, then turn usingReRank on and off to see the ranking of the top few items; switch searchMode and adjust limit, record the number of candidates, scores, and cited text tokens, and record the array item count separately.

Business test: use one real record to find fields inconsistent with the ledger, mark discrepancies in name, quantity, date, or signature, and submit them to the reviewer for confirmation and write-back.

  • In the knowledge base search test, switch searchMode to semantic, full-text, and hybrid, compare the number of candidates and similarity scores, and check the fields, dates, and signatures of review invitations, expert review comments, and author revision notes.
  • Turn usingReRank on and off to compare the ranking of the top few items, adjust limit to observe changes in the number of candidates, check version synchronization and receipt signatures, and record cited text tokens and array item count.