What this category's data looks like
This category's data typically includes laboratory grading and classification tables, hazard source lists, and safety responsibility statements. Sources are departmental submissions, laboratory self-inspections, and summaries by school management departments. The update cadence is mostly centralized revision each academic year or semester, with sporadic supplementary entries during the period. Field characteristics center on names, quantities, dates, and signature fields: the grading and classification table records laboratory names, category levels, and responsible persons; the hazard source list records hazard source names, quantities, and storage locations; the safety responsibility statement records responsible persons and signing dates. In multilingual materials and terminology processing scenarios, when these three types of materials enter knowledge base retrieval and recall, each plays a different role: the grading and classification table provides laboratory entries and level terminology; the hazard source list provides hazard source names and quantifier expressions; the safety responsibility statement provides responsible person titles and date formats. The terminology list and comparison draft need to extract consistent wording from these materials, then refer back to the original text.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate snippets and similarity scores (after mode, merge, re-rank, filter, and upper-limit trimming). "Unified terminology list and comparison draft" is completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Laboratory grading and classification table | Names, category levels, responsible persons, and dates are self-consistent within the same record, and consistent with corresponding records in the hazard source list and safety responsibility statement | In the search test, search by laboratory name, check the name and level wording in candidate snippets, record the number of candidates and scores, and use searchMode to switch between semantic, full-text, or hybrid modes to compare recall |
| Hazard source list | Hazard source names, quantities, and storage locations can refer back to each other between the list and the grading and classification table, and quantity unit wording is unified | Extract three hazard source terms, check whether the wording is consistent throughout, check whether candidate snippets include both semantic and full-text results, and record the number of candidates before and after similarity filtering |
| Safety responsibility statement | Responsible person titles, signing dates, and receipt status are complete, and consistent with the responsible person in the grading and classification table | Check the signing date and responsible person wording, view the order of the top few candidates before and after enabling usingReRank, and confirm that quoted text tokens and array counts are recorded separately |
| Retrieval return results | Candidate snippets and similarity scores are returned after mode, merge, re-rank, filter, and upper-limit trimming, and quoted text retains key snippets from the original text | Test once each with synonym-rewritten questions, compare candidates and scores, and check the number of quoted items and content completeness after limit trimming |
Three common errors
- Phenomenon: The name or quantity of the same laboratory does not match between two materials. Cause: Different reporting standards, wording not unified in the terminology list, and retrieval candidates each take one version.
- Phenomenon: The grading and classification table has been updated, but the hazard source list is still the old version. Cause: Versions are not synchronized, retrieval hits old snippets, and quoted content is inconsistent with the current materials.
- Phenomenon: The safety responsibility statement lacks a signature or receipt. Cause: The signing process is not closed-loop, and retrieval can only return candidate snippets in an incomplete state.
How to verify the setup
Acceptance of this feature: Test once each with synonym-rewritten questions, compare candidates and similarity scores, then toggle re-ranking on and off to observe the ordering of the top few; record the number of candidates and scores, and check the effects of searchMode, similarity, usingReRank, and limit.
Business test: Extract three terms and check whether the wording is consistent throughout, and whether the translated names can refer back to the original text.
- In the knowledge base search test, switch retrieval modes, compare candidate results, and record the number of candidates and scores.
- Check the tokens of the quoted text after upper-limit trimming, record the array count separately, and confirm that key snippets are retained.