What this category's data looks like
This category's data looks like this: Continuing education credit recognition involves three types of materials: course filing forms, study certificates, and credit recognition forms. Fields include course name, credit hours, study date, signature, and receipt. The course filing form is used to verify the course name and filed credit hours; the study certificate is used to verify the actual study date and quantity; the credit recognition form is used to verify the recognition conclusion and signature. Materials are updated weekly. The same record often does not match across two materials in terms of name, quantity, or date, and versions are not synchronized or signatures/receipts are missing. After the three types of materials enter the knowledge base retrieval service, each serves as a source of candidate segments, split by fields and dates, for subsequent nodes or humans to generate progress tables and statistical scope explanations.
Configuration and implementation checks
Configuration and implementation checks: The knowledge base retrieval service (retrieval node and search test) only outputs candidate segments and similarity scores (after pattern, merge, rerank, filter, and limit trimming). Progress tables and statistical scope explanations must be produced by subsequent nodes or humans. A single retrieval consists of question optimization, semantic/full-text/hybrid recall, multi-path merging, rerank secondary sorting, similarity filtering, and reference limit trimming, and only then is the reference content handed to the model.
| Object | Check | Verification action |
|---|---|---|
| Course filing form | Whether the course name and filed credit hours fields are consistent with the study certificate, and whether the date and signature are complete | In the knowledge base search test, switch searchMode to semantic, full-text, or hybrid, compare whether the candidate segments cover both the course name and credit hours fields, and record the number of candidates and scores |
| Study certificate | Whether the actual study date and credit hours quantity match the credit recognition form, and whether the receipt is missing | Test once each with synonym-rewritten questions, check whether the candidates include the date and quantity fields, use similarity to filter low-scoring segments, and separately record the number of items in the reference array |
| Credit recognition form | Whether the recognition conclusion and signature correspond to the course filing form and study certificate, and whether the versions are synchronized | Compare the ranking of the top few items before and after enabling usingReRank, check whether the recognition conclusion field ranks first, use limit to control the number of candidates, check the reference text tokens, and record the array item count separately |
| Progress table and statistical scope explanation | Whether the number of records and the statistical count match, whether the scope is written in the header, and whether exceptions are clearly stated | Use one week of records to verify the total, check whether the scope and exceptions are clearly stated, and check whether the candidate segments retain the key original text segments |
Three common errors
- Symptom: The course name, credit hours quantity, and study date of the same record do not match between the course filing form and the study certificate. Cause: The two materials were entered at different times, versions are not synchronized, and fields are not aligned.
- Symptom: The credit recognition form lacks a signature or receipt, and the recognition conclusion cannot correspond to the study certificate. Cause: Signatures were missed during material circulation, receipts were not archived, and versions are not synchronized.
- Symptom: The same question cannot be recalled when rephrased, and although there are many candidates, the ones ranked at the top are irrelevant. Cause: The retrieval mode does not match the material form, rerank is not enabled, and similarity filtering and reference limit trimming are not properly tuned.
How to verify the setup
Feature acceptance: Test once each with synonym-rewritten questions, compare candidates and scores, then toggle rerank to observe the ranking of the top few items; switch searchMode to see whether semantic and full-text results are merged, use similarity to filter low-scoring segments, use limit to control the number of candidates, and separately record the number of items in the reference array.
Business testing: Use one week of records to verify the total, check whether the scope and exceptions are clearly stated, check whether the number of records and the statistical count match, and whether the scope is written in the header.
- In the knowledge base search test, switch
searchMode, compare the candidate segments and scores of the semantic, full-text, and hybrid modes, and check whether the course name, credit hours quantity, and study date fields are complete. - Compare the ranking of the top few items before and after enabling
usingReRank, usesimilarityto filter low-scoring segments, uselimitto control the number of candidates, and record the reference text tokens and array item count separately.