What this category's data looks like
This category's data covers three types of materials: written and interview registration forms, Mandarin proficiency certificates, and medical examination forms. Sources are exam authority announcements, certification body notices, and applicant submissions. The update cadence changes with batches and policy adjustments. Fields include name, ID number, education, exam subjects, scores, dates, signatures, and receipts. When materials enter knowledge base retrieval and recall, their roles differ: the registration form provides eligibility and subject information, the Mandarin certificate provides grade and number, and the medical examination form provides conclusions and physician signatures. For the same record, the name, quantity, and date in the registration form and the medical examination form must be consistent. If versions are not synchronized or signatures/receipts are missing, it directly affects the certification decision.
Configuration and implementation checks
This step is performed 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 version ledger and change records are completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Written and interview registration form | Name, quantity, date, and signature must be consistent with the medical examination form and Mandarin certificate | In the search test, test once each with synonymous questions, record the number of candidates and scores, and check searchMode and similarity |
| Mandarin certificate | Grade, number, and issue date must match the registration form record | Switch retrieval modes to compare candidates, enable usingReRank to see changes in the ranking of the top few results |
| Medical examination form | Conclusion, date, and physician signature are complete; missing signature or receipt is considered inconsistent | Check the tokens of the cited text and the number of items after limit trimming, and confirm that key passages are not lost |
| Version ledger | The current version is unique, old versions are traceable, and citations are updated synchronously | After changing one piece of content, check whether the version number and citations are synchronized, and verify candidates and scores |
Three common errors
- Symptom: The name, quantity, and date in the registration form and the medical examination form for the same record do not match. Cause: Material versions are not synchronized, and the old version is still being recalled.
- Symptom: The Mandarin certificate number is inconsistent with the registration form record. Cause: The knowledge base was not updated after the change, and citations still point to the old snippet.
- Symptom: The medical examination form lacks a physician signature or receipt. Cause: The uploaded document is incomplete, and retrieval candidates do not cover the signature page.
How to verify the setup
Acceptance for this feature: Test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank on and off to see the ranking of the top few results.
Business test: Change one piece of content and see whether the version number and citations are updated synchronously.
- In the search test, switch
searchModeand compare the number of candidates and scores for semantic, full-text, and hybrid modes. - Compare the order of the top few results before and after enabling
usingReRank, and check the cited content aftersimilarityandlimittrimming.