What this category's data looks like
This category's data looks like this: Technical college school-enterprise education management involves three core types of materials: student registration forms, teaching plans, and skills training records. Student registration forms are updated by the admissions and student records department according to semester registration nodes, recording student identity, major, class, and registration date; teaching plans are compiled by the academic affairs department according to the professional talent training program, including course names, class hours, semesters, and approval signatures; skills training records are filled in weekly by the training workshop, including training items, workstations, quantities, and instructor signatures. The three types of materials serve as mutual evidence in fields, dates, and signatures. When entering the knowledge base retrieval and recall function, student registration forms serve as identity baseline fragments, teaching plans serve as course and class-hour reference fragments, and skills training records serve as process and quantity corroborating fragments, jointly supporting consistency verification and discrepancy location of names, quantities, and dates for the same object across two materials.
Configuration and implementation checks
Configuration and implementation checks: This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate fragments and similarity scores (after pattern, merge, rerank, filter, and limit trimming). The discrepancy list is completed subsequently by downstream nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Student registration form | Compare with the teaching plan and skills training records to check whether the same student's name, major, and registration date are consistent | In the search test, retrieve with synonymous questions, switch searchMode to semantic, full-text, and hybrid, and record the number of candidates and scores |
| Teaching plan | Compare with the student registration form and skills training records to check course names, class-hour quantities, course start dates, and approval signatures | Enable and disable usingReRank, compare the order of the top few candidates, and check whether key fields are retained after similarity filtering |
| Skills training record | Compare with the teaching plan and student registration form to check training items, workstation quantities, training dates, and instructor signatures | Adjust limit to observe changes in the number of candidates, and record referenced text tokens and array item counts separately |
| Discrepancy list | The same object's name, quantity, and date are consistent across the two materials, and signatures and receipts are complete | Backfill field by field according to candidate fragments, submit to the reviewer for confirmation, and write back to the ledger |
Three common errors
- Phenomenon: The same student's name or student ID does not match between the student registration form and the skills training record. Cause: The two materials were not associated with the same object during data entry, and retrieval candidates are returned independently.
- Phenomenon: The course quantity or training dates in the teaching plan and skills training record are inconsistent. Cause: Versions are not synchronized, and old-version fragments remain in the knowledge base and participate in recall.
- Phenomenon: The skills training record lacks an instructor signature or receipt, but retrieval still returns candidate fragments. Cause: Candidates and referenced content are not authorization results, and signature status requires manual verification.
How to verify the setup
How to verify the setup: Acceptance of this function: test once each with synonym-rewritten questions, compare candidates and similarity scores, then toggle rerank on and off to inspect the ordering of the top few results.
Business test: use one real record to find fields that are inconsistent with the ledger.
- In the search test, fix one real record and retrieve separately in semantic, full-text, and hybrid modes, recording the number of candidates and scores.
- Inspect the top few candidates once before and once after enabling rerank, and check whether the tokens of the referenced text retain key fields after
similarityfiltering andlimittrimming.