🎯 Diderot Article Dashboard - 6 Pathway System

📅 Last Updated: 2025-12-10

Welcome to the Diderot Article Dashboard!

This intelligent news analysis system helps you filter and score articles using advanced pathway analysis and composite scoring algorithms.

🔧 Basic Controls

• Time Interval: Controls how far back to search for articles (15 min to 90 days)
  • Smart Time Limits: Maximum time window automatically adjusts based on filter specificity:
  • • Domain filter active → 90 days max (most specific)
  • • Specific User + Specific Topic → 90 days max (ultra-focused)
  • • Specific Topic → 60 days max (topic research)
  • • Specific User → 30 days max (user stream)
  • • All Users + All Topics → 10 days max (broadest query)
  • ⚠️ Time automatically reduces when switching to broader filters (with warning message)
  • 💡 Pro Tip: Open multiple browser tabs for different time windows/filters simultaneously
• 🔄 Update Button: Refreshes article analysis and re-runs pathway calculations
• Pause/Resume: Dashboard starts paused - click Resume to begin loading articles
  • Auto-resumes when computer wakes from sleep (if not manually paused)
• Domain Filter: Type domain name to filter articles at the database level for accurate counts and better performance
  • Enables 90-day time window for deep domain analysis
• Sort Order: Choose how articles are sorted - by insertion date or publication date, newest or oldest first

🎯 Advanced Filtering

• User: Filter by specific content analysts or show all
• Articles: Show all articles, only included ones, or only excluded ones
• Pathway: Filter by pathway results (ALL, IN-ALL, OUT-ALL, or specific pathways)
• Composite: Filter by composite score results (Comp-IN, Comp-OUT) based on threshold setting
• Display Nulls: When unchecked (default), hides DR=null articles from display. Metrics still count them if they meet other criteria
• Acceptability: Filter by metric acceptability thresholds
• Category & Topic: Filter by content category and specific topic/deck

📊 Pathway Analysis System

• Six Pathways: Configure up to 6 different analysis pathways with custom thresholds
• Logic Types: Use AND (all criteria must pass) or OR (any criteria can pass)
• Criteria: Domain Reliability, Semantic Score, BM25, KNA Score, and ES Score
• Results: Articles classified as IN (passes pathway) or OUT (fails pathway)
• Save/Load: Save pathway configurations as named sets for reuse

🎯 Composite Scoring & Settings

• Composite Score: A 0-100 score combining DR (via multiplier), Semantic (42.5%), BM25 (42.5%), and KNA (15%) metrics using weighted average
• Composite Score Settings: Configure scoring behavior:
  • Ignore DR Null: When checked, articles with DR=null get scored (DR treated as neutral). When unchecked, they cannot pass composite scoring
  • Apply Pathway Bonus: Articles passing all 3 active pathways get +20% bonus, failing all get -20% penalty
  • Apply DR Multiplier: DR acts as score multiplier (see DR Multiplier Logic below)
  • DR Min: Minimum raw DR value required (0-5 scale, 0 = no minimum)
  • Max DR Multiplier: Multiplier applied at DR=100 score (default 1.5x, means 50% boost)
  • Min Sem: Minimum semantic score required (0-1 scale, 0.475 default, 0 = no minimum)
  • Composite Threshold: Pass/fail cutoff score (default: 55)
  • 💾 Save Button: Saves all composite settings to browser localStorage
• DR Multiplier Logic: Domain Reliability acts as a score multiplier:
  • DR below minimum → 0.5x multiplier (50% penalty)
  • DR at minimum threshold → 1.0x multiplier (baseline, no change)
  • DR at maximum (100 score) → configurable multiplier (default 1.5x)
  • Linear interpolation between minimum and maximum
  • Example: With DR Min=2.5 (score 50) and Max=1.5x, a DR=4.0 (score 75) gets 1.25x multiplier
• COMPOSITE PASSES: Summary line shows count and pass rate of articles meeting threshold
• New Composite Score Changes: IN-C (passes composite but is_excluded=true) and OUT-C (fails composite but is_excluded=false)
• MISMATCH-COMP Badge: Articles showing disagreement between composite score and current production status

📈 Statistics & Analysis

• Summary Stats: View mean, median, std dev for each scoring approach
• Pass Rates: See how many articles pass each pathway
• Statistics Breakdown: Five key summary lines:
  • OLD SYSTEM: Shows counts based on the is_excluded database field - the current production inclusion/exclusion decision
  • NEW PATHWAYS: Shows how many articles pass each of the 6 configurable pathways
  • ACTIVE PATHWAYS: Shows enabled pathway count, new pass rate (% passing at least one pathway), and old pass rate (% with is_excluded=false)
  • UNIQUE PATHWAY PASSES: Breaks down articles by which specific pathway(s) they pass
  • COMPOSITE PASSES: Shows articles passing the composite score threshold with pass rate %
• Active Decks: Shows unique feed/deck counts from all loaded articles (not just filtered view)
• Future-Dated Articles: Tracks articles where published_at > feed_inserted_at (impossible dates). Hover to see top domains and average delay times
• Batch History: Click "Articles Loaded" metric to see last 20 batch loads with timestamps and counts
• Distribution Charts: Analyze delay patterns and domain reliability distributions
• Copy Functions: Export article data and analysis results
• Performance Limits: Dashboard loads max 1,500 articles per fetch to maintain browser performance

🔍 Understanding Article Cards

• Color Coding: Blue (originally included), Gray (originally excluded)
• Status Badges: Shows old system result vs new pathway result
• MISMATCH-PATH Badge: Article's pathway result differs from production is_excluded status
• MISMATCH-COMP Badge: Article's composite score result differs from production is_excluded status
• Pathway Results: Green checkmarks show which pathways passed
• Composite Score: Shows 0-100 composite score value when available
• Metrics: DR, Semantic, KNA, BM25, ES scores with color coding

💡 Pro Tip: Start by configuring pathways, then use the various filters (User, Inclusion, Pathway, Category, Topic) to find articles that meet your criteria. The 6-Pathway system provides comprehensive article analysis.

Overview: The composite scoring system combines multiple article quality metrics into unified 0-100 scores using three different algorithmic approaches. This allows for comprehensive article evaluation and comparison across different methodologies.

Approach 1: Threshold-Weighted Scoring

• Method: Uses hard thresholds with weighted component scoring
• DR Filter: Articles with DR < 3.0 get harsh penalties (score × 0.3)
• Component Weights: SEM (40%), BM25 (35%), KNA (25%)
• Pathway Bonus: Each passed pathway adds multiplier (1.1× per pathway)
• Best For: Clear quality thresholds, binary decision making

Approach 2: Data-Driven Percentile

• Method: Normalizes scores against dataset distribution using percentiles
• Dynamic Scaling: Converts raw scores to percentile ranks (0-100)
• DR Treatment: Harsh penalties for unreliable sources (DR < 3), favorable scaling for reliable sources (DR ≥ 3)
• Component Weights: DR (25%), SEM (30%), BM25 (30%), KNA (15%)
• Pathway Bonus: Additive bonus (+1.05× per pathway)
• Best For: Relative ranking, harsh DR penalties, avoiding score inflation

Approach 3: Multi-Stage Filtering

• Method: Uses DR as a multiplier with additive component bonuses
• DR Multiplier: Domain reliability acts as overall quality multiplier
• Additive Scoring: SEM, BM25, and KNA contribute additive bonuses
• Pathway Bonus: Exponential bonus (15 × 1.5^pathways)
• Best For: Emphasizing source quality while rewarding content metrics

Overview: The KNA Entity system adjusts article scores based on how many defined keyword entities (aliases) are found in the article. This rewards articles that mention all key entities and penalizes those missing key entities.

Advanced Quality Filter: Articles are filtered using a comprehensive multi-tier exclusion system with DR floor requirements, dual-low correlation detection, and enhanced penalty structures.

📰 Example 1: High-Quality Reliable Source

Article: DR=4.2, SEM=0.75, BM25=12.5, KNA=0.85, Pathways=2

⚠️ Example 2: Poor Quality Unreliable Source

Article: DR=2.1, SEM=0.75, BM25=12.5, KNA=0.85, Pathways=2

🚫 Example 3: Quality-Based Exclusions

🔍 Key Insights

• DR Impact: Sources with DR < 3 receive harsh penalties across all approaches, dramatically reducing scores
• Quality Exclusions: Articles with BM25=0, SEM=0, or missing KNA are excluded unless from trusted sources (DR ≥ 4)
• Trusted Source Grace: High-reliability sources (DR ≥ 4) get proportional reweighting when metrics are missing
• Approach Differences: Each approach emphasizes different aspects - use multiple approaches for comprehensive evaluation

💡 Usage Tip: Compare all three approaches to understand different perspectives on article quality. Use Approach 1 for strict quality gates, Approach 2 for relative rankings, and Approach 3 when source reliability is paramount.