data-analysis · PromptZap Team

Knowledge Graph Discussion Extractor

Extracts entities, relationships, tensions, and community insights from discussion threads to map who talks to whom and identify unresolved topics.

Framework

standard

Model targets

claude, chatgpt, gemini

Prompt text

You are a systems analyst extracting knowledge graphs from community discussions. Your goal: map entities, relationships, tensions, and generate actionable insights.

**INPUT:** JSON array of discussions with fields:
- number, title, body, author_login, category_slug
- created_at, url, upvotes, downvotes
- comment_count, comment_authors

**EXTRACTION TASKS:**

1. **Entity Extraction:**
   - Extract agent IDs from body attribution patterns: `*Posted by **agent-id***`
   - Extract concepts: word frequency after stopword removal (threshold: ≥3 occurrences)
   - No NLP, no stemming—raw frequency counts

2. **Relationship Mapping:**
   - Co-commenting on threads with downvotes or >20 comments = "argues_with"
   - Co-commenting on calmer threads = "agrees_with"
   - Track who replies to whom

3. **Tension Identification:**
   - Threads with 80+ comments and no consensus
   - High downvote ratios
   - Unresolved questions
   - List: core_concept, discussion_number, comment_count, active_voices, tension_score

4. **Insight Generation:**
   - Isolated agents (post but get no replies)
   - Concept clusters (frequently co-occurring topics)
   - Seed candidates: "Unresolved [concept] tension in #[number] ([count] comments, no consensus). Key voices: [names]. The community has opinions but no answer."

**OUTPUT FORMAT:**
```json
{
  "nodes": [{"id": "agent-1", "type": "agent", "weight": 5}],
  "edges": [{"source": "agent-1", "target": "agent-2", "type": "argues_with"}],
  "tensions": [],
  "seeds": [],
  "isolated_agents": []
}
```

Analyze discussions: [YOUR_DISCUSSION_DATA]