Run2 document metadata extraction

Document Metadata Extraction

Overview

This skill covers extracting metadata (author, date, type, reviewers, feedback) from documents and correlating them with discussion threads and review feedback.

Document Structure Analysis

Identifying Document Metadata

def analyze_document_structure(doc):
    """Extract all metadata from a document"""
    metadata = {
        'type': doc.get('type'),           # e.g., 'Technical Specifications'
        'id': doc.get('id'),              # e.g., 'contentforce_tech_spec'
        'author': doc.get('author'),      # e.g., 'eid_8efef28a'
        'date': doc.get('date'),          # Document creation date
        'feedback': doc.get('feedback'),  # Review feedback (often text)
        'document_link': doc.get('document_link'),  # URL to document
        'content_preview': doc.get('content', '')[:200],  # First 200 chars
    }
    return metadata

Understanding Document Feedback Format

def extract_feedback_insights(feedback):
    """Parse feedback string for reviewer comments"""
    if isinstance(feedback, str):
        # Feedback is often a bulleted list of suggestions
        lines = feedback.split('\n')
        suggestions = [line.strip('- ').strip() for line in lines if line.strip()]
        return suggestions
    elif isinstance(feedback, dict):
        # Sometimes structured as dict with reviewers
        return feedback.get('comments', [])
    elif isinstance(feedback, list):
        return feedback
    return []

Identifying Reviewers

Strategy 1: Slack Document Share Discussions

def find_reviewers_in_slack(slack_msgs, doc_id, doc_type):
    """Find who reviewed/commented on a document in Slack"""
    reviewers = set()

    for msg in slack_msgs:
        msg_text = msg.get('Message', {}).get('User', {}).get('text', '')
        msg_user = msg.get('Message', {}).get('User', {}).get('userId', '')

        # Find the message where document was shared
        if doc_id in msg_text or doc_type.lower() in msg_text.lower():
            if 'shared' in msg_text.lower() or 'shared the' in msg_text.lower():
                # This is the author sharing the document
                author = msg_user

                # Now look for thread replies - these are the reviewers
                for reply in msg.get('ThreadReplies', []):
                    reply_user = reply.get('userId', '')
                    if reply_user and reply_user != 'slack_admin_bot' and reply_user != author:
                        reviewers.add(reply_user)

    return list(reviewers)

Strategy 2: Updated Document Notifications

def extract_reviewers_from_updates(slack_msgs, doc_id):
    """When document is updated with "thank you everyone", find recent participants"""
    reviewers = set()
    thank_you_msgs = []

    # Find "thank you everyone" messages
    for i, msg in enumerate(slack_msgs):
        msg_text = msg.get('Message', {}).get('User', {}).get('text', '')
        if 'thank you everyone' in msg_text.lower() and doc_id in msg_text:
            thank_you_msgs.append(i)

    # Look back from these messages to find earlier discussions
    for thank_you_idx in thank_you_msgs:
        # Scan backwards to find the original share message
        for j in range(thank_you_idx - 1, max(0, thank_you_idx - 50), -1):
            msg = slack_msgs[j]
            msg_text = msg.get('Message', {}).get('User', {}).get('text', '')
            msg_user = msg.get('Message', {}).get('User', {}).get('userId', '')

            if doc_id in msg_text and 'shared' in msg_text.lower():
                # Found the original share, extract thread participants
                for reply in msg.get('ThreadReplies', []):
                    reply_user = reply.get('userId', '')
                    if reply_user and reply_user != 'slack_admin_bot':
                        reviewers.add(reply_user)
                break

    return list(reviewers)

Strategy 3: Meeting Participation

def find_reviewers_in_meetings(meetings, document_type):
    """Find reviewers from meeting attendance on document-related meetings"""
    reviewers = set()

    for meeting in meetings:
        meeting_id = meeting.get('id', '').lower()

        # Check if meeting relates to the document type
        # e.g., 'product_dev_ContentForce' meeting for ContentForce documents
        if 'contentforce' in meeting_id and 'product_dev' in meeting_id:
            participants = meeting.get('participants', [])
            reviewers.update(participants)

    return list(reviewers)

Document Type Handling

Distinguishing Document Versions

def identify_document_versions(docs):
    """Find original and updated versions of documents"""
    documents = {}

    for doc in docs:
        doc_id = doc.get('id', '')

        # Separate original from final versions
        if 'final_' in doc_id:
            base_id = doc_id.replace('final_', '')
            if base_id not in documents:
                documents[base_id] = {}
            documents[base_id]['final'] = doc
        else:
            if doc_id not in documents:
                documents[doc_id] = {}
            documents[doc_id]['original'] = doc

    return documents

Correlating Versions

def correlate_document_versions(docs, slack_msgs):
    """Find both original and updated versions, track reviewers"""
    result = {}

    versioned = identify_document_versions(docs)

    for base_id, versions in versioned.items():
        if 'original' in versions:
            author = versions['original'].get('author')

            # Find reviewers from Slack discussion about original
            reviewers = find_reviewers_in_slack(slack_msgs, base_id, '')

            result[base_id] = {
                'author': author,
                'original_version': versions.get('original'),
                'final_version': versions.get('final'),
                'reviewers': reviewers
            }

    return result

Practical Workflow

def analyze_documents_completely(docs, slack_data, meetings):
    """Complete analysis pipeline"""
    results = {}

    # Step 1: Get all document metadata
    for doc in docs:
        doc_id = doc.get('id')
        if 'final_' not in doc_id:  # Skip final versions initially
            results[doc_id] = {
                'metadata': analyze_document_structure(doc),
                'author': doc.get('author'),
            }

    # Step 2: Find reviewers through multiple strategies
    for doc_id, data in results.items():
        # Try Slack discussions
        slack_reviewers = find_reviewers_in_slack(slack_data, doc_id, data['metadata']['type'])

        # Try meeting participation
        meeting_reviewers = find_reviewers_in_meetings(meetings, data['metadata']['type'])

        # Combine, remove author from reviewers
        all_reviewers = list(set(slack_reviewers + meeting_reviewers))
        author = data['author']
        data['reviewers'] = [r for r in all_reviewers if r != author]

    return results

Tips

• Document IDs are key - Use them to correlate across Slack and other sources
• Multiple versions - Always check for both original and "final_" versions
• Feedback is metadata - The feedback field often contains reviewer suggestions
• Slack share dates - Look for the specific share message to find review threads
• Meeting attendance - Cross-check with meeting participants for validation
• Remove duplicates - Use sets to deduplicate across different discovery methods