What this category's data looks like
This category's data covers written exam and interview registration forms, Mandarin proficiency certificates, medical examination forms, and other materials. Sources are exam institutions and certification institutions, and updates advance with batches and acceptance cycles. Registration forms carry the exam subjects, test districts, and payment status; Mandarin proficiency certificates carry the grade and number; medical examination forms carry the conclusion and physician signature. When materials enter knowledge base retrieval and recall, each participates in recall as an independent fragment: registration forms are used to verify exam eligibility and subjects, Mandarin proficiency certificates are used to verify grade and validity period, and medical examination forms are used to verify the conclusion and signature receipt. Field characteristics center on name, ID number, date, quantity, and signature fields. The name, date, and quantity of the same record across different materials need to be aligned item by item. Authorized access and external services require that visitor identity match the visible scope: parents or superiors only see the portion of materials they are authorized to see, and unauthorized content is not visible.
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 mode, merge, rerank, filter, and upper-limit trimming). The externally available entry points and authorization scope are then completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Written exam and interview registration form | Name, ID number, exam subjects, test district, payment status, and signature field are aligned item by item; dates and quantities are consistent across the two materials | In the search test, switch searchMode, compare the candidate counts and scores of the semantic, full-text, and hybrid modes, and check whether the registration form fragments are hit |
| Mandarin proficiency certificate | Grade, number, issue date, and validity period fields are complete, and match the name, date, and quantity in the registration form record | Test once each with synonym-rewritten questions, observe changes in candidates and scores, and enable usingReRank to compare the ordering of the top few results |
| Medical examination form | Conclusion, examination date, physician signature, and receipt are complete, and the version is synchronized with the registration form | After filtering by similarity, view the candidate count and scores, record the tokens and array count of the cited text, and check them separately |
| Restricted visitor identity | Identity matches the visible scope, and unauthorized materials are not visible | Access once with a restricted identity and see whether the returned candidates fall only within the authorized scope, with unauthorized materials not appearing |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match between the registration form and the medical examination form. Cause: The two materials are entered separately without field alignment, so they are hit independently during retrieval recall.
- Symptom: The Mandarin proficiency certificate version is not synchronized, and the old grade coexists with the new registration form record. Cause: After the material is updated, the associated fragment is not written back, so two versions appear among the candidates at the same time.
- Symptom: The medical examination form lacks a signature or receipt but is still recalled as a valid fragment. Cause: The signature field is not included in the filter conditions, and similarity filtering only trims by text relevance.
How to verify the setup
Feature acceptance: test once each with synonym-rewritten questions, compare candidates and scores, then toggle rerank on and off to inspect the ordering of the top few results; switch searchMode and similarity, usingReRank, and record the candidate count and scores.
Business testing: access once with a restricted identity, see whether the visible scope matches the setting, and whether unauthorized materials are not visible.
- In the knowledge base search test, switch
searchMode, compare the candidate counts and scores of the semantic, full-text, and hybrid modes, and confirm that the material form matches the mode. - Compare the ordering of the top few results before and after enabling
usingReRank, and after filtering bysimilarity, record the tokens and array count of the cited text, and check them separately.