What this category's data looks like
Kindergarten operation behavior and quality self-assessment materials usually include self-assessment reports, problem lists, rectification plans, as well as accompanying observation records, sign-in sheets, receipts, and signature pages. The materials are updated by semester or year, and the fields involve kindergarten name, class, date, item name, quantity, responsible person, and signature and receipt status. When entering knowledge base retrieval and recall, the self-assessment report serves as the overall basis and source of conclusions, the problem list serves as the itemized problems and rectification direction, the rectification plan serves as the time nodes and responsibility division, and the signature and receipt pages serve as procedural evidence. Each material enters the candidate pool in the form of fragments, and retrieval results need to return to the corresponding position in the original text before they can be used to answer parents or be written into a plan.
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 upper-limit trimming). The basis items that can point back to the original text are then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Self-assessment report | The kindergarten name, class, date, and signature page in the report must be aligned item by item with the problem list and rectification plan, and the name and quantity of the same item must not drift between the two materials | In the search test, search by item name, check whether the candidate fragments hit both the report and the list, record the number of candidates and scores, and compare the date and signature fields |
| Problem list | The problem description, involved class, and discovery date of the list items are consistent with the corresponding items in the rectification plan, the version should be the current one, and old drafts must not participate in recall | Search once each with synonymous questions, switch searchMode to semantic, full-text, and hybrid, check whether the candidates cover the original text of the list, and verify the version date |
| Rectification plan | The rectification items, responsible persons, and completion dates in the plan match the problem list and receipt records, and items missing signatures or receipts need to be marked separately | Compare the ranking of the top few items before and after enabling usingReRank, check whether the plan items are stably ranked high, and separately record the tokens and array count of the quoted text |
| Retrieval results | Candidate fragments can point back to the specific material and original text position, key sentences are still retained after similarity filtering, and upper-limit trimming of citations only affects the number of items, not source attribution | Adjust similarity and limit to observe changes in the number of candidates and scores, and confirm that the quoted text has not been truncated to the point of losing its basis meaning |
Three common errors
- Symptom: The name, quantity, or date of the same record does not match between the self-assessment report and the problem list. Cause: The two materials are maintained separately, without field-level alignment.
- Symptom: The list item referenced by the rectification plan has already been modified in the current version. Cause: The version was not synchronized, and the old draft is still in the knowledge base participating in recall.
- Symptom: The signature page or receipt is missing, but retrieval still returns that item as a basis. Cause: Procedural evidence was not included in the candidate filtering conditions.
How to verify the setup
Acceptance for this feature: test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank to inspect the ranking of the top few items, and verify the return differences under searchMode, similarity, limit, and usingReRank.
Business test: ask a basis question that spans materials and see whether the answer gives a traceable source.
- In the knowledge base search test, retrieve the same item using two phrasings, record the number of candidates and scores, and point back to the original text position item by item.
- Compare the order of the top few items before and after enabling rerank, separately record the tokens and array count of the quoted text, and verify whether the signature and receipt fields are complete.