[Feature Suggestion] Add Dedicated "Health & Wellness" and "Movies & Cinema" Verticals to Explore & Topic Filtering

[SB2318]

Summary

Adding dedicated Health & Wellness and Movies & Cinema vertical categories to the X Explore section and topic recommendation pipeline.

Currently, recommendations rely on broader topic buckets and post-based filtering (via TopicFilteringId in phoenix_topics_source.rs). Introducing distinct Explore tabs and topic hydrators for Health and Movies will allow the algorithm to surface highly relevant, domain-specific content to users while unlocking significant ad revenue and engagement opportunities.

---

🌐 Product & Business Justification

1. "Health & Wellness" Vertical

• Business Impact: High-value advertiser ecosystem (pharmaceuticals, fitness brands, supplements, wearables, and digital health services) seeking targeted ad placement.
• Creator Monetization: Empowers verified medical professionals, fitness coaches, and health researchers to build dedicated subscriber bases.
• User Retention & Engagement: Health updates, wellness tips, and research discussions drive daily recurring user habits and high intent engagement.

2. "Movies & Cinema" Vertical

• Business Impact: Prime monetization opportunity for entertainment studios, streaming platforms, and movie theater chains launching trailer campaigns and premiere promotions.
• Real-Time Cultural Moments: X is already the global town square for live reaction conversations (box office releases, movie trailers, film festivals, and award shows).
• Rich Media Optimization: Directly leverages video engagement scores (video_open_score, vqv_score) for trailer previews and behind-the-scenes clips.

---

🛠️ Proposed Technical Implementation Plan

1. Topic Source & Hydration:
   • Extend phoenix_topics_source.rs and filtered_topics_hydrator.rs to include topic models for HEALTH_WELLNESS and MOVIES_CINEMA.
2. Parameter Configurations:
   • Configure feature flags in param.rs (TopicFilteringOverrides, TopicOonWeightFactor) to fine-tune out-of-network recommendation multipliers for health insights and movie trailer media.
3. Scoring Adjustments:
   • Tailor RankingScorer weighting for media-rich movie posts (boosting video_open_weight and photo_expand_weight) and high-reply/discussion health threads (reply_weight).

---

📊 Success Metrics & KPIs

• Explore Tab Engagement: Increased daily active users (DAU) and dwell time within topic feeds.
• Ad Monetization Yield: Higher eCPM rates for health and movie industry ad placements.
• Media Engagement: Increased video view counts (vqv_score) and bookmark rates for movie trailers and health educational content.

[SB2318]

---

Technical Feasibility & Architecture Breakdown

1. Topic Taxonomy Expansion

The repository uses a topic taxonomy system defined in home-mixer/filters/topic_ids_filter.rs.

• How it works: TopicIdExpansion maps topic category IDs to subtopics and supertopics.
• Change Required: Register new Topic Category IDs for HEALTH_WELLNESS and MOVIES_CINEMA in TopicIdExpansion.

2. Candidate Retrieval Source

Topic candidate retrieval is handled by PhoenixTopicsSource.

• How it works: When a user requests a topic feed, PhoenixTopicsSource dispatches requests to the Phoenix ML inference cluster using topic IDs (query.topic_ids).
• Change Required: Pass the newly registered Health and Movie topic IDs when dispatching retrieval requests.

3. Topic Hydration & Filtering

Topic tags are attached to posts via FilteredTopicsHydrator and filtered by TopicIdsFilter.

• How it works: Candidates fetched from retrieval clusters are filtered and verified against candidate topic IDs (c.filtered_topic_ids).
• Change Required: Ensure topic entity classifiers tag candidate posts with the new Health and Movie topic entity IDs.

4. Custom Scoring Weights per Vertical

Scoring and weighting are driven by RankingScorer and parameters in param.rs.

• How it works: The algorithm already supports topic-specific out-of-network weighting (TopicOonWeightFactor) and override rules (TopicFilteringOverrides).
• Customization Options:
  • Movies Vertical: Boost video_open_weight and photo_expand_weight to favor trailers and visual previews.
  • Health Vertical: Boost reply_weight and follow_author_weight to prioritize expert discussions and trusted health creators.

---

Summary

Because the pipeline already abstracts topics via TopicIdExpansion, PhoenixTopicsSource, and TopicIdsFilter, introducing Health and Movies feeds is a clean, native expansion of the recommendation engine without requiring architectural rewrites.