Knowledge Base Retrieval for Training Provider Scheduling and Lesson Records: Trace Exceptions and Review Results

Training Schedules: review retrieval for exception review. Check source versions, input mappings and the values received by downstream nodes.

What this category's data looks like

This page addresses the class scheduling and class-hour deduction scenario for training institutions. The materials include three types: class schedules, class-hour deduction records, and class-change applications. Class schedules record the class, course, teacher, classroom, and start/end dates; class-hour deduction records record the student, number of class hours, deduction date, and handler; class-change applications record the original class session, new class session, applicant, and approval signature. The three types of materials are updated weekly or monthly, field names are often inconsistent between systems and paper forms, and date formats and class-hour quantity conventions also differ. When the materials enter knowledge base retrieval, each is recalled as an independent fragment: class schedule fragments are used to verify scheduling dates and classrooms, class-hour deduction record fragments are used to verify class-hour quantities and signatures, and class-change application fragments are used to verify the class sessions before and after the change and the approval receipt. The retrieval service only returns candidate fragments and similarity scores; the exception list and review conclusions are generated by subsequent nodes or manually based on these.

Configuration and implementation checks

This step is performed by the knowledge base retrieval service (retrieval node and search test), outputting candidate fragments and similarity scores (after mode, merge, rerank, filter, and limit trimming). The exception list and review conclusions are then completed by subsequent nodes or manually.

ObjectCheckVerification action
Class scheduleCheck whether the class, course, teacher, classroom, and start/end dates are consistent with the class-hour deduction recordsIn the search test, switch searchMode to semantic, full-text, or hybrid, search once each with the class schedule field names and aliases, and record the number of candidates and scores
Class-hour deduction recordCheck whether the student, number of class hours, deduction date, and handler signature are consistent with the class-change applicationSearch using the combination of student and date from the class-hour deduction record, and compare the ranking changes of the top few candidates after enabling and disabling usingReRank
Class-change applicationCheck whether the original class session, new class session, applicant, and approval signature are consistent with the class scheduleSearch separately using the original class session and new class session from the class-change application, and check whether the candidates include both semantic and full-text recall results
Retrieval parametersCheck whether the candidates retain key fragments of the original text after setting similarity and limitAdjust similarity and limit and search once each, record the number of candidates and the tokens of the cited text, and record the array count separately

Three common errors

  • Symptom: The course name and number of class hours for the same class session do not match between the class schedule and the class-hour deduction record. Cause: The two materials were filled in by different handlers, and the field conventions and update rhythms are not synchronized.
  • Symptom: The class-change application has been approved, but the class schedule and class-hour deduction record are still the old version. Cause: The versions were not synchronized, and the class-change receipt was not written back to the class schedule and class-hour deduction record.
  • Symptom: The class-hour deduction record lacks the handler signature, or the class-change application lacks the approval receipt. Cause: Signatures were missed when supplementally entering paper materials, or the electronic process was archived before completion.

How to verify the setup

Feature acceptance: test once each with synonym-rewritten questions, compare candidates and scores, then toggle rerank on and off to observe the ranking of the top few; check the number of candidates and scores, and verify the effects of searchMode, usingReRank, similarity, and limit.

Business test: create an exception record, see whether the source material can be traced, and verify the names, quantities, dates, and signatures in the class schedule, class-hour deduction record, and class-change application.

  • In the knowledge base search test, switch searchMode, search once each with the field names and aliases from the class schedule, class-hour deduction record, and class-change application, and record the number of candidates and scores.
  • Compare the ranking of the top few candidates after enabling and disabling usingReRank, verify the tokens of the cited text after adjusting similarity and limit, and record the array count separately.