A graph-based media intelligence system for exploring creative networks across Chinese and American film industries.
Curated demo dataset: 204 people in CSV, 338 people nodes in the interactive graph, 135 films, 3,715 relationships.
Most media tools treat films, creators, awards, genres, and platforms as separate tables. MediaGraph models the industry as a connected network so you can see how talent, taste, recognition, and distribution shape strategic opportunities.
The goal is to feel like an internal strategy product for a streaming platform, studio, talent agency, or entertainment investor.
- Maps creators, films, genres, awards, platforms, countries, languages, decades, audience segments, and themes in one graph
- Surfaces bridge figures between Chinese-language cinema and American cinema
- Scores director-actor creative fit with explainable sub-scores
- Shows genre opportunity maps by country, decade, awards, and platform
- Generates graph-grounded strategy summaries from retrieved network facts
- Supports an optional Neo4j workflow with reusable Cypher examples
The opening hero examples now point to public references in data/public_sources.md, and the repo includes a source map at data/public_source_map.csv that attaches public URLs to people, films, and awards.
- Featured bridge creators: Ang Lee, Chloé Zhao, Michelle Yeoh
- Featured bridge films: Crouching Tiger, Hidden Dragon; Everything Everywhere All at Once
- Official award databases are linked for the major awards / recognition layer
The broader CSV graph is still a curated demo dataset, but this adds a concrete verified source trail for the public-facing story.
A polished dashboard summary with metric cards for people, films, relationships, Chinese-language creators, American creators, and the most central / bridge-like nodes.
An interactive network graph with filters for:
- country or region
- role
- genre
- decade
- award status
- market type
- platform
Select a person and inspect:
- profile summary
- known works
- collaborators
- genres, awards, and platforms
- centrality and cross-market signals
- similar talents
- ego network visualization
A transparent prototype score built from:
- genre overlap
- thematic compatibility
- award pathway similarity
- collaboration proximity
- audience compatibility
- cross-market potential
Shows which creators, countries, awards, and platforms cluster around a genre.
Highlights creators and films that connect Chinese and American markets using:
- betweenness centrality
- degree centrality
- cross-market score
- bridge films and bridge creators
A rule-based development assistant for project ideas. It returns:
- recommended genres
- reference films
- potential directors
- potential actors
- likely platforms
- audience segments
- market risks
- a positioning statement
Shows graph facts first, then a concise explanation built from those facts. The goal is transparency, not hype.
Surfaces creators and films connected to major awards and festival pathways.
Summarizes the most central creators, bridge creators, cross-market films, award-connected creators, and promising pairings.
Nodes:
- Person
- Film
- Genre
- Award
- Platform
- Country
- Language
- Decade
- AudienceSegment
- Theme
Relationships:
- DIRECTED
- ACTED_IN
- WROTE
- PRODUCED
- BELONGS_TO_GENRE
- WON_AWARD
- NOMINATED_FOR
- DISTRIBUTED_ON
- RELEASED_IN
- MADE_IN_COUNTRY
- USES_LANGUAGE
- HAS_THEME
- APPEALS_TO
- SIMILAR_TO
- COLLABORATED_WITH
- CROSSES_MARKET
The score is a prototype, not a predictive model. It combines:
- genre overlap between director and actor filmographies
- thematic overlap
- award pathway similarity
- collaboration proximity in the graph
- audience compatibility through platform footprint
- cross-market potential based on the creator profiles
Each sub-score is visible in the app so the logic stays explainable.
MediaGraph identifies bridge nodes by combining:
- betweenness centrality
- degree centrality
- cross-market score from the creator profile
- bridge-film signals from mixed-market titles
It is especially useful for spotting creators and projects that can travel between Chinese-language markets and the U.S. market.
The app first retrieves graph facts, such as:
- connected films
- awards and festival signals
- cluster membership
- cross-market links
It then turns those facts into a short strategy explanation. The point is to make the reasoning visible before the narrative.
- Python
- Streamlit
- Pandas
- NetworkX
- Plotly
- Neo4j optional
- scikit-learn optional for future similarity work
pip install -r requirements.txt
streamlit run app.py- Start Neo4j locally or in Aura.
- Load the schema in
graph/neo4j_schema.cypher. - Use
graph/sample_queries.cypherto explore bridge creators, awards, and genre clusters. - Expand the demo CSVs with verified public data if you want a larger production-ready graph.
mediagraph/
├── app.py
├── requirements.txt
├── README.md
├── data/
│ ├── people.csv
│ ├── films.csv
│ ├── roles.csv
│ ├── awards.csv
│ ├── platforms.csv
│ ├── audience_signals.csv
│ └── edges.csv
├── graph/
│ ├── neo4j_schema.cypher
│ └── sample_queries.cypher
├── assets/
│ └── screenshots/
└── utils/
├── graph_builder.py
├── scoring.py
└── insights.py
- Which directors and actors are most central in Chinese and American film networks?
- Which creators frequently cross genres, countries, or platforms?
- Which directors have strong award pathways?
- Which actor-director combinations look strategically promising?
- What kinds of films connect Chinese and American entertainment markets?
- Which creators are bridge figures between Chinese-language cinema and American cinema?
- The dataset is a curated demonstration dataset, not a complete industry database.
- Some sample fields are synthetic or generalized when verified public data was not necessary for the prototype.
- The README and app intentionally avoid claiming prediction accuracy.
- The project does not scrape private data or use non-public personal information.
MediaGraph demonstrates:
- graph database thinking
- connected data modeling
- media strategy and talent discovery
- recommendation logic and scoring design
- explainable insight generation
- cross-cultural entertainment analysis
- interactive visualization and data storytelling
- product-oriented UX for a recruiter-facing portfolio
MediaGraph is designed as a two-layer portfolio product:
- Web experience:
web/contains a Vercel-ready static site with draggable bubble-network exploration and a people-only mode. - Analyst view:
app.pyruns the deeper Streamlit strategy workspace with fit scoring, bridge analysis, and graph explanations.
- GitHub repo: source of truth for code, data, schema, and documentation
- Vercel: deploy the
web/directory as the public product-facing site - Streamlit Community Cloud: deploy
app.pyas the analyst console