What this category's data looks like
District and county teaching research activity and classroom observation and evaluation materials come from three types of sources: teaching research activity notices, classroom observation records, and classroom evaluation minutes, updated weekly or monthly. Notices contain the activity name, date, subject, participating schools, and signature fields; classroom observation records contain the instructor, class, lesson topic, date, observation points, and signature; classroom evaluation minutes contain evaluation points, conclusions, acknowledgments, and signatures. When the materials enter knowledge base retrieval and recall, notices carry the activity scope and boundary definitions, classroom observation records carry process facts and date anchors, and classroom evaluation minutes carry conclusions and exception explanations. The three types of materials serve as mutual verification objects for fields, dates, and signatures; the name, quantity, and date of the same record in two materials must be aligned item by item. When versions are not synchronized or signatures or acknowledgments are missing, the statistical scope explanation must state the exception handling method in the table header, so that subsequent nodes or humans can generate the progress table and statistical scope explanation.
Configuration and implementation checks
This step is executed by the knowledge base retrieval service (retrieval node and search test), outputting candidate chunks and similarity scores (after mode, merge, rerank, filter, and limit truncation). The "progress table and statistical scope explanation" is completed by subsequent nodes or humans.
| Object | Check | Verification action |
|---|---|---|
| Teaching research activity notice | Whether the name, date, subject, participating schools, and signature fields are consistent with the classroom observation records | After switching searchMode in the knowledge base search test, test once each with synonym-rewritten questions, record the number of candidates and scores, and check whether the names and dates of the notices and classroom observation records are aligned |
| Classroom observation record | Whether the instructor, class, lesson topic, date, and signature are consistent with the classroom evaluation minutes | Turn usingReRank on and off to compare the ranking of the first few items, check whether the name, quantity, and date of the same record match across the two materials, and mark missing signatures or acknowledgments separately |
| Classroom evaluation minutes | Whether the evaluation points, conclusions, acknowledgments, and version are synchronized with the notices and classroom observation records | Adjust similarity to observe changes in the number of candidates and scores, check whether the conclusions of the classroom evaluation minutes cover the scope of the notices, and write exception explanations in the table header where versions are not synchronized |
| Cited content | After truncation, key original text fragments are still retained; the number of items and the budget are recorded separately | Record the number of items in the returned citation array, separately check whether the tokens of the cited text are within limit and the citation budget, and do not compare the array length with the citation budget for equality |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match between the notice and the classroom observation record. Cause: The two materials were entered separately, versions are not synchronized, and signatures or acknowledgments are missing.
- Symptom: The conclusions of the classroom evaluation minutes are inconsistent with the scope of the notices, and exceptions appear in the statistical totals. Cause: The minutes do not cover all activities, and missing acknowledgments mean the scope is not clearly stated in the table header.
- Symptom: There are many candidates, but the ones ranked first 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 loose.
How to verify the setup
This feature acceptance: test once each with synonym-rewritten questions, compare candidates and scores, then turn rerank on and off to inspect the ranking of the first few items; switch searchMode, usingReRank, and similarity to observe changes in the number of candidates and scores.
Business test: use one week of records to check the totals and see whether the scope and exceptions are clearly stated; verify the fields, dates, and signatures of notices, classroom observation records, and classroom evaluation minutes item by item, and confirm that the record counts and statistical counts match.
- In the knowledge base search test, switch
searchMode, test once each with synonym-rewritten questions, record the number of candidates and scores, and check whether the names and dates of the notices and classroom observation records are aligned. - Turn
usingReRankon and off to compare the ranking of the first few items, adjustsimilarityto observe candidate changes, record the number of items in the citation array and separately check the tokens of the cited text, and write exceptions into the statistical scope table header.