What this category's data looks like
For multilingual materials and terminology processing in vocational college skills competition organization, the materials include competition regulations, registration forms, training plans, as well as a unified terminology glossary and a comparative draft. The regulations are published by the organizer, registration forms are submitted by each participating team, training plans are updated monthly by the instructors, and the terminology glossary is synchronized on a rolling basis with revisions to the regulations. The fields involve project name, group, date, signature, receipt, quantity, language, and translated name. When entering the knowledge base retrieval service, the regulations and training plans serve as terminology context and sentence pattern references, the registration forms provide name and date comparisons, and the terminology glossary and comparative draft serve as anchors for translated names referring back to the source text. The update rhythm is mainly centralized revisions before the competition and temporary supplementary entries during the competition; versions easily diverge, and signatures and receipts are often delayed.
Configuration and implementation checks
This stage is executed 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 unified terminology glossary and comparative draft are completed by subsequent nodes or manually.
| Object | Check | Verification action |
|---|---|---|
| Competition regulations | Whether the fields, dates, and signatures match those in the registration forms and training plans, and whether terminology usage is consistent throughout | In the search test, switch searchMode, search once each with different translations of the same term, and check the number of candidates and scores |
| Registration forms | Whether the name, quantity, and date of the same record are consistent in the regulations and training plans, and whether receipts are complete | Compare whether the returned candidates contain both semantic and full-text results, record the number of candidates and scores, and check the date and signature fields |
| Training plans | Whether the version is synchronized with the regulations, and whether translated terminology can refer back to the source text | With usingReRank enabled and disabled, observe the order of the top few results each time, and check the impact of similarity and limit on candidates |
| Unified terminology glossary and comparative draft | Whether the same term is written consistently throughout, and whether translated names can refer back to the source text | Sample three terms to check their usage throughout, check the tokens of the quoted text, and record the number of array items separately |
Three common errors
- Phenomenon: The name and date of the same project do not match in two materials. Cause: The registration form and regulation versions are not synchronized, and the terminology glossary has not been updated on a rolling basis with revisions to the regulations.
- Phenomenon: Training plans lack signatures or receipts, and translated terminology is inconsistent. Cause: Temporary supplementary entries during the competition were not backfilled, and no unified proofreading was done after version divergence.
- Phenomenon: There are many candidates, but the ones ranked at the front are all irrelevant, and the answer lacks key paragraphs. 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
Acceptance for this feature: test once each with synonymously rewritten questions, compare candidates and scores, then toggle rerank on and off to observe the ranking of the top few results; check the impact of similarity, limit, searchMode, and usingReRank on the number of candidates and scores, and record quoted text tokens and the number of array items separately.
Business test: sample three terms to check whether their usage is consistent throughout.
- In the knowledge base search test, switch modes, compare candidate results, and confirm that after multi-path merging both semantic and full-text results are included.
- Compare the order of the top few results before and after enabling rerank, and check whether key source snippets are still retained after similarity filtering and quoted limit trimming.