What this category's data looks like
This category's data looks like this: Compulsory education dropout prevention and control involves three core types of materials: dropout risk ledgers, persuasion-to-return records, and home visit records. Fields include student name, grade, risk level, persuasion-to-return date, home visit time, signature, and receipt. Material sources are schools, townships, and county-level education departments, and the update cadence changes with the semester, persuasion-to-return progress, and home visit frequency. When these three types of materials enter knowledge base retrieval and recall, each plays a different role: the dropout risk ledger serves as the primary index anchor, providing students' basic information and risk level; persuasion-to-return records serve as process evidence, supplementing the number and results of persuasion-to-return attempts; home visit records serve as supporting materials, supplementing family communication and signed receipts. During retrieval, matching must be performed by material type and date fields to ensure the current version is unique, old versions are traceable, and citations are synchronized.
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 chunks and similarity scores (after mode, merge, rerank, filter, and limit trimming). The "version ledger and change records" are completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Dropout risk ledger | Fields include student name, grade, risk level, date, and signature; the same record's name, quantity, and date must be consistent across two materials | In the knowledge base search test, switch searchMode to semantic, full-text, or hybrid, compare the number of candidates and scores, and check whether the ledger fields match the persuasion-to-return records |
| Persuasion-to-return records | The number of persuasion-to-return attempts, dates, signatures, and receipts must be synchronized with the ledger; when the version is not synchronized, the old version is traceable | Test once each with synonymous questions, compare the ranking of the top few results before and after enabling usingReRank, and check whether the dates and signatures in the persuasion-to-return records and home visit records are consistent |
| Home visit records | Home visit time, signature, and receipt must be consistent with the persuasion-to-return records; when signature or receipt is missing, it must be marked | Adjust similarity and limit, observe changes in the number of candidates and scores, and check whether the name, quantity, and date in the home visit records match the ledger |
| Version ledger and change records | The current version is unique, old versions are traceable, and citations are synchronized; change one piece of content and see whether the version number and citations update synchronously | In the search test, change one piece of content, observe whether the candidate chunks and scores update synchronously, record the number of citations and citation text tokens, and record the array count separately |
Three common errors
- Symptom: The same student's name and date do not match between the ledger and the persuasion-to-return records. Cause: Field conventions were not unified when the materials were entered, the version was not synchronized, and the old version is still being recalled.
- Symptom: Home visit records lack signatures or receipts, and the cited content is incomplete. Cause: Signatures and receipts were not verified when the materials were archived, and retrieval was not filtered by field.
- Symptom: The same question phrased differently cannot be recalled, and although there are many candidates, the ones ranked at the top are all irrelevant. Cause: The retrieval mode does not match the material form, rerank is not enabled, or similarity filtering is too strict.
How to verify the setup
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 inspect the ranking of the top few results; switch searchMode and usingReRank, and record the number of candidates and scores.
Business test: change one piece of content and see whether the version number and citations update synchronously; adjust similarity and limit, verify citation text tokens, and record the array count separately.
- In the knowledge base search test, switch
searchModeto semantic, full-text, or hybrid, compare the number of candidates and scores, and check the fields, dates, and signatures of the ledger, persuasion-to-return records, and home visit records. - Compare the ranking of the top few results before and after enabling
usingReRank, adjustsimilarityandlimit, verify citation text tokens, record the array count separately, and confirm that the current version is unique and old versions are traceable.