Skip to main content

OJT Parser

Table of Contents​

  1. Overview
  2. Getting Started
  3. Input Data Requirements
  4. How to Prompt the Agent
  5. Example Usage
  6. Best Practices
  7. Troubleshooting
  8. FAQ

Overview​

The OJT Data Parser Agent is a specialized agent designed to extract and parse structured information from On-the-Job Training (OJT) documents. It focuses on extracting two main categories of information:

  • Project (OJT) Summary Information: Comprehensive metadata about the training program
  • Project (OJT) Curriculum Tasks: Detailed task breakdowns including steps, standards, and learning objectives

Key Capabilities​

  • Extracts structured data from various document formats
  • Parses complex OJT task analysis tables
  • Generates CSV files for easy data import and manipulation
  • Displays extracted data in readable markdown table format
  • Maintains data integrity by only extracting existing information

Getting Started​

Prerequisites​

  • Access to OJT training documents (PDF, Word, or text formats)

Input Data Requirements​

Document Format​

The agent accepts documents containing:

  1. Training Module Information

    • Module title and descriptions
    • Company and location details
    • Personnel assignments (approver, assignee, lead)
    • Timeline information

  2. Task Analysis Tables

    • Main task listings
    • OJT hours allocations
    • Task breakdown structures including:
      • Task steps
      • Key learning points
      • Standards and requirements
      • Required skills and knowledge
      • Training guidelines

Expected Structure​

Documents should ideally contain:

  • Clear section headers (e.g., "MODULE", "MAIN TASKS", "OJT HOURS")
  • Structured tables for task analysis
  • Consistent formatting for dates and durations

How to Prompt the Agent​

Basic Prompt Structure​

Please extract OJT information from the uploaded document [filename]

Advanced Prompting Tips​

  1. Specific Extraction: Request focus on particular sections

    Extract only the curriculum tasks from the OJT document

  2. Multiple Documents: Process batch documents

    Parse all OJT modules in the uploaded training materials

  3. Validation Requests: Verify extraction completeness

    Extract and validate all task analysis tables from the document

Example Usage​

Example 1: Complete OJT Document Extraction​

Input Document Content:

MODULE: Equipment Maintenance Training
COMPANY: TechCorp Industries
LOCATION: Singapore
APPROVER: John Smith
ASSIGNEE: Jane Doe
START DATE: 2024-01-15
END DATE: 2024-03-15

MAIN TASKS:
1. Preventive Maintenance - 40 hours
2. Troubleshooting - 60 hours

ON-THE-JOB TRAINING TASK ANALYSIS:
S/N || Main Task || Steps || Learning Points || Standards
1 || Preventive Maintenance || Check oil levels || Understanding viscosity || ISO 9001
2 || Troubleshooting || Diagnose issues || Problem identification || Company SOP

Expected Output:

  • Project Summary CSV with all metadata fields
  • Project Tasks CSV with formatted task descriptions
  • Markdown tables for immediate viewing

Example 2: Task-Focused Extraction​

Prompt:

Extract only the curriculum tasks with their detailed breakdown from the OJT document

Output Focus:

  • Detailed task_description_table with proper CSV formatting
  • Preserved table structure with || delimiters
  • All columns and rows maintained

Best Practices​

Document Preparation​

  1. Ensure Clear Structure: Use consistent headers and formatting
  2. Complete Tables: Include all columns even if empty
  3. Date Formatting: Use YYYY-MM-DD format for dates
  4. Duration Units: Clearly specify hours, days, or weeks

Prompt Optimization​

  1. Be Specific: Clearly state what information you need extracted
  2. Provide Context: Mention document type and expected content
  3. Request Validation: Ask for confirmation of extracted fields

Data Handling​

  1. Review Output: Always verify extracted data against source
  2. Check CSV Format: Ensure proper delimiter usage (semicolon)
  3. Validate Completeness: Confirm all expected fields are present

Troubleshooting​

Common Issues and Solutions​

Issue 1: Missing Fields in Output​

Symptom: Some expected fields are empty in the CSV Solution:

  • Verify the field exists in the source document
  • Check for alternative field names or variations
  • Ensure document quality is sufficient for parsing

Issue 2: Malformed Task Description Tables​

Symptom: Task analysis tables are not properly formatted Solution:

  • Ensure source tables use consistent delimiters
  • Check for merged cells or irregular formatting
  • Manually clean the source document if necessary

Issue 3: CSV Export Fails​

Symptom: Generated CSV files cannot be opened or imported Solution:

  • Verify delimiter consistency (semicolon)
  • Check for unescaped quotes in data
  • Ensure proper line endings

Issue 4: Incomplete Extraction​

Symptom: Agent misses entire sections of the document Solution:

  • Break document into smaller sections
  • Ensure clear section headers
  • Remove complex formatting or graphics

FAQ​

Q1: What document formats are supported?​

A: The agent primarily works with text-based formats including PDF, Word documents, and plain text files. Complex layouts or image-heavy documents may require preprocessing.

Q2: Can the agent handle multiple OJT modules in one document?​

A: Yes, the agent can extract multiple modules. Each will be listed as a separate row in the Project Summary CSV, with corresponding tasks in the Tasks CSV.

Q3: How does the agent handle missing information?​

A: The agent leaves fields empty when information is not found. It does not generate or infer missing data, maintaining data integrity.

Q4: What is the maximum document size the agent can process?​

A: While there's no strict limit, very large documents (>100 pages) may benefit from being split into sections for optimal processing.

Q5: Can I customize the output format?​

A: The agent outputs in CSV format with semicolon delimiters and markdown tables. The CSV structure is fixed but can be post-processed as needed.

Q6: How accurate is the extraction?​

A: Accuracy depends on document quality and structure. Well-formatted documents typically achieve 95%+ accuracy. Always review critical data.

Q7: Can the agent extract from scanned documents?​

A: Scanned documents require OCR preprocessing. The agent works best with digital text rather than image-based content.

Q8: How are complex tables with merged cells handled?​

A: The agent attempts to preserve table structure but merged cells may cause formatting issues. Consider reformatting complex tables before extraction.

Q9: Is there support for non-English documents?​

A: The agent is optimized for English documents. Other languages may work but with reduced accuracy.

Q10: Can I extract specific date ranges or filter results?​

A: The agent extracts all available data. Filtering should be done post-extraction using the CSV outputs.