What this category's data looks like
The materials for self-funded study abroad application services fall into three tracks: the study abroad service contract, the school application form, and the personal statement. The contract provides the scope of engagement, fee items, and the signing dates of both parties; the application form provides the institution name, program direction, application round, and submission deadline; the personal statement provides narrative paragraphs, word count, and the finalized version. The three materials each play different roles in the knowledge base retrieval and recall stage: the contract is the source of authorization and responsibility boundaries, the application form is the source of institution and time fields, and the personal statement is the source of text fragments and semantic recall. Materials are updated by application season, and before each round's submission the finalized version is replaced and a receipt is countersigned. Field characteristics center on four categories: names, quantities, dates, and signatures. The same record often shows inconsistencies in name spelling, date conventions, and version numbers between the contract and the application form, so the candidate fragments and similarity scores returned by retrieval must be used as the starting point for verification.
Configuration and implementation checks
The knowledge base retrieval service (retrieval node and search test) only outputs candidate fragments and similarity scores (after pattern, merge, re-rank, filter, and limit trimming). The handover-ready deployment state and operation records must be continued by subsequent nodes or manually. One retrieval consists of question optimization, semantic/full-text/hybrid recall, multi-path merging, re-rank secondary sorting, similarity filtering, and reference limit trimming, and only then is the reference content handed to the model obtained.
| Object | Check | Verification action |
|---|---|---|
| Study abroad service contract | The contract fields must cover the scope of engagement, fee items, the signing dates of both parties, and the signature receipt; the name and date of the same record should be consistent between the contract and the application form | In the knowledge base search test, retrieve with a contract clause question, check whether the returned candidate fragments contain the signing date and signature receipt, record the number of candidates and the score, and check the value of searchMode |
| School application form | The application form fields must cover the institution name, program direction, application round, and submission deadline; the version number must be consistent with the finalized personal statement | Retrieve with an institution name and round question, compare whether the semantic and full-text candidates hit at the same time, record the number of candidates and the score, and check whether the deadline field is still retained after similarity filtering |
| Personal statement | The personal statement must cover narrative paragraphs, word count, and the finalized version; the name and quantity of the same paragraph should match across the two materials | Retrieve with a personal statement paragraph question, compare the order of the top few items before and after enabling re-rank, record the number of candidates and the score, and check the reference text tokens after usingReRank and limit trimming |
Three common errors
- Symptom: The name and date of the same record do not match between the contract and the application form. Cause: The two materials were finalized separately, versions were not synchronized, and the retrieval candidates hit only one of them.
- Symptom: The personal statement word count is inconsistent with the quantity filled in the application form. Cause: After the finalized version was replaced, the receipt was not countersigned, and the old version still participates in recall in the knowledge base.
- Symptom: The answer lacks key paragraphs. Cause: After similarity filtering and reference limit trimming, candidate fragments are truncated; the reference text tokens and array count need to be checked.
How to verify the setup
Acceptance of this feature: Test once each with synonym-rewritten questions, compare candidate fragments and similarity scores, then toggle re-rank on and off to observe the ordering of the top few items; record the number of candidates and the score, and check the actual values of the three identifiers searchMode, similarity, and usingReRank in the search test.
Business test: Check item by item once according to the handover sheet to see which record is missing; against the fields, dates, and signatures of the contract, application form, and personal statement, confirm that the version, connections, and task records are consistent.
- In the knowledge base search test, switch retrieval modes, compare the number of candidates and the score, and confirm that the candidates include both semantic and full-text results.
- Check item by item the names, quantities, dates, and signature receipts of the contract, application form, and personal statement; record missing items in the handover sheet and countersign.