What this category's data looks like
This page addresses the scenario of primary and secondary school exam question setting and grading. The materials include three types: question-setting two-way specification tables, exam papers, and grading rubrics. The sources are electronic documents submitted by teaching and research departments and schools, and the update cadence rolls with exam batches and revised versions. When the three types of materials enter the document parsing and chunking function, their roles differ: the specification table carries the mapping of assessment points and question types, the exam paper carries the question and score structure, and the grading rubric carries the answer key points and scoring criteria. Field characteristics are concentrated in question numbers, assessment point numbers, scores, dates, and signature fields. Multilingual materials also involve terminology in the original language and translated names for comparison. Parsing and chunking cuts each material into retrievable data rows. Terminology and translated names are scattered across different materials and need to be consolidated into a unified terminology table and comparison draft in the subsequent retrieval step.
Configuration and implementation checks
This step is executed by parsing and training tasks (server-side tasks), outputting data rows and index status (parsed, indexing, indexed). The "unified terminology table and comparison draft" is then completed by retrieval and subsequent processing.
| Object | Check | Verification action |
|---|---|---|
| Question-setting two-way specification table | Assessment point numbers, question types, and scores must correspond item by item with the exam paper and grading rubric, and date and signature fields must be complete | Check whether the question numbers, assessment point numbers, and scores in the specification table and exam paper are consistent, and check dates and signatures; after parsing and chunking, use chunkSize and chunkSplitter to see whether the chunk boundaries fall between question numbers and assessment points |
| Exam paper | If the name, quantity, or date of the same record does not match across the two materials, it is judged inconsistent | Check the file records and data rows separately to confirm that the identities of the file records and data rows are distinguishable; check whether the data row status is parsed, indexing, or indexed, and verify that the interface display is consistent |
| Grading rubric | Answer key points and scoring criteria must correspond to the exam paper's question numbers and scores, and the version must be synchronized with the specification table | Sample three terms and check whether their spelling is consistent throughout the whole document, and verify whether translated names can refer back to the original text; check the data rows covered by the rebuild task to confirm that only data with completed indexing is covered |
| Unified terminology table and comparison draft | Only when the same term is consistent throughout the whole document and translated names can refer back to the original text can it be released externally | In the retrieval results, distinguish processing modes by trainingType, check terminology entries and original text sources; record the number of cited entries and the tokens of the cited text, and register the two separately |
Three common errors
- Phenomenon: The name, quantity, or date of the same record does not match between the specification table and the grading rubric. Cause: The versions of the two materials are not synchronized; only one of them was updated in the revised draft, and after parsing and chunking they became independent data rows.
- Phenomenon: The exam paper lacks a signature or receipt, making it impossible to trace responsibility for question setting and review after release. Cause: The signature field was not submitted together with the electronic document; parsing and chunking retained only the main text, and the signature page was not included in the same data row.
- Phenomenon: Newly uploaded materials cannot be found, or empty data rows remain after a training failure. Cause: The data row is in parsed or indexing status, and data pending indexing does not participate in recall; failed training data has not been edited or deleted.
How to verify the setup
Acceptance for this feature: upload one material and observe the data row status transitions and the coverage scope of the rebuild task; confirm the statuses are parsed, indexing, and indexed in sequence. Changing models only rebuilds data with completed indexing, and the data rows covered by the rebuild task do not include data pending indexing.
Business test: sample three terms and check whether their spelling is consistent throughout the whole document, verify whether translated names can refer back to the original text, and record the number of cited entries and the tokens of the cited text.
- Open the data rows to check status, confirm that it is in one of the three states and consistent with the interface display; check the file records and data rows separately to confirm that the identities of the file records and data rows are distinguishable.
- Check the data rows covered by the rebuild task to confirm that only data with completed indexing is covered; use
chunkSize,chunkSplitter, andtrainingTypeto check chunking and processing modes.