Knowledge Base Retrieval for College Entrance Exam Application Eligibility: Handle Multilingual Materials and Terminology

What this category's data looks like

Under the general college entrance examination registration qualification review scenario, materials include the college entrance examination registration form, household registration and student status certificates, materials for children accompanying migrant workers, as well as multilingual versions of terminology glossaries and reference drafts. The registration form records the candidate's basic information and examination subjects; the household registration and student status certificates carry the household registration address, student status start and end dates, and school seal; the materials for children accompanying migrant workers include fields such as residence permit, social insurance contributions, and guardian relationship. The material sources are electronic files submitted by admissions and examination institutions at all levels and schools, and the update cadence rolls forward as registration batches progress; field characteristics are concentrated in names, ID numbers, dates, quantities, signatures, and receipts. When entering knowledge base retrieval, the registration form and certificate-type materials serve as candidate segment sources, the terminology glossary and reference drafts serve as the basis for back-referencing translated names, and the multilingual versions are used to verify whether the writing of the same term is consistent throughout the entire text.

Configuration and implementation checks

This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate segments and similarity scores (after pattern, merge, re-rank, filter, and upper-limit trimming). Unifying the terminology glossary and reference drafts is then completed by subsequent nodes or manually.

Three common errors

• Phenomenon: The same candidate's name is written differently in the registration form and the household registration and student status certificates. Cause: The two materials were submitted in different batches, the terminology glossary was not updated synchronously, and retrieval candidates were not aligned across materials.
• Phenomenon: The number of social insurance months in the materials for children accompanying migrant workers does not match the quantity filled in the registration form. Cause: Versions were not synchronized, old-version materials still participated in recall in the knowledge base, and date filtering was not applied after merging.
• Phenomenon: The household registration and student status certificates lack a signature or receipt. Cause: The signature page was omitted when uploading the scanned copy, key paragraphs were trimmed after similarity filtering, and candidate segments cannot back-reference the original text.

How to verify the setup

This feature acceptance: Test once each with synonymously rewritten question sentences, compare candidates and scores, then toggle re-ranking on and off to observe the ordering of the top few; record the number of candidates and scores, check the tokens of the cited text, and record the array length separately.

Business test: Sample three terms to check whether their writing is consistent throughout the entire text, check whether the translated names can back-reference the original text, and confirm that the candidate segments of the same term in the multilingual versions come from the same recall channel.

• In the knowledge base search test, switch searchMode, record the number of candidates and scores under the semantic, full-text, and hybrid modes respectively, and check whether the material fields are complete.
• Compare the order of the top few candidates before and after enabling usingReRank, and check whether the cited text retains the key segments of the original text after trimming by similarity and limit.