What this category's data looks like
The materials on this page come from pre-departure training and overseas safety reminder scenarios, including three types: pre-departure training courseware, safety reminder checklists, and accommodation contact forms. The courseware contains course modules, terminology definitions, and multilingual comparisons; the checklist contains risk items, emergency steps, and a receipt column; the contact form contains accommodation address, contact person, date, and signature fields. The materials are mainly updated periodically, with versions adjusted according to batches and destinations. When entering knowledge base retrieval and recall, the courseware and checklist serve as sources for semantic and full-text recall, and the contact form serves as a source of structured fields. In the same retrieval request, the three respectively contribute candidate chunks for subsequent nodes or manual work to generate a unified terminology glossary and comparison draft.
Configuration and implementation checks
The knowledge base retrieval service (retrieval node and search test) only outputs candidate chunks and similarity scores (after pattern, merge, rerank, filter, and upper limit trimming). The unified terminology glossary and comparison draft must be done by subsequent nodes or manually. One retrieval consists of question optimization, semantic/full-text/hybrid recall, multi-route merging, rerank secondary sorting, similarity filtering, and citation upper limit trimming, and only then is the citation content handed to the model obtained.
| Object | Check | Verification action |
|---|---|---|
| Pre-departure training courseware | The same term is written consistently throughout the whole text, and the translated name can refer back to the original text | In the search test, test once each with synonymous questions, compare candidates and scores; switch searchMode to observe the differences among semantic, full-text, and hybrid options |
| Safety reminder checklist | The name, quantity, and date of the same record do not match across the two materials | Check whether the candidate chunks simultaneously contain semantic and full-text results; enable usingReRank to compare the ordering of the top few items |
| Accommodation contact form | The version is not synchronized; signature or receipt is missing | Record the tokens of the cited text and the number of array items separately; adjust similarity to see whether key original text chunks are still retained after filtering |
Three common errors
- Phenomenon: The name, quantity, and date of the same record do not match between the courseware and the checklist. Cause: The versions of the two materials are not synchronized, and during recall each hits chunks from different batches.
- Phenomenon: The accommodation contact form lacks a signature or receipt column. Cause: The material update cycle is disconnected from the publishing process, and the candidate chunks do not cover the receipt field.
- Phenomenon: The same question cannot be recalled when rephrased. Cause: Question optimization does not cover references and expansion, or
searchModedoes not match the material form.
How to verify the setup
Acceptance of this feature: test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank on and off to observe the ordering of the top few items.
Business test: sample three terms and check whether their writing is consistent throughout the whole text.
- In the knowledge base search test, switch
searchMode, compare candidate results and scores, and record the number of array items. - Before and after enabling
usingReRank, observe the order of the top few items once each, then adjustsimilarityandlimitto check the tokens of the cited text.