Skip to content

nevesliu7/mediagraph

Repository files navigation

MediaGraph

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.

Live demo · Dashboard

Why I built this

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.

What it does

  • 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

Verified public source coverage

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.

Key features

1) Opening intelligence brief

A polished dashboard summary with metric cards for people, films, relationships, Chinese-language creators, American creators, and the most central / bridge-like nodes.

2) Global creative network

An interactive network graph with filters for:

  • country or region
  • role
  • genre
  • decade
  • award status
  • market type
  • platform

3) Talent network explorer

Select a person and inspect:

  • profile summary
  • known works
  • collaborators
  • genres, awards, and platforms
  • centrality and cross-market signals
  • similar talents
  • ego network visualization

4) Director-actor creative fit score

A transparent prototype score built from:

  • genre overlap
  • thematic compatibility
  • award pathway similarity
  • collaboration proximity
  • audience compatibility
  • cross-market potential

5) Genre opportunity map

Shows which creators, countries, awards, and platforms cluster around a genre.

6) Cross-market bridge analysis

Highlights creators and films that connect Chinese and American markets using:

  • betweenness centrality
  • degree centrality
  • cross-market score
  • bridge films and bridge creators

7) Story positioning assistant

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

8) Graph-based explanation demo

Shows graph facts first, then a concise explanation built from those facts. The goal is transparency, not hype.

9) Awards and prestige pathway

Surfaces creators and films connected to major awards and festival pathways.

10) Strategic insight dashboard

Summarizes the most central creators, bridge creators, cross-market films, award-connected creators, and promising pairings.

Data model

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

How the Creative Fit Score works

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.

How the Cross-Market Bridge Analysis works

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.

How the graph-based explanation demo works

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.

Tech stack

  • Python
  • Streamlit
  • Pandas
  • NetworkX
  • Plotly
  • Neo4j optional
  • scikit-learn optional for future similarity work

How to run locally

pip install -r requirements.txt
streamlit run app.py

Optional Neo4j setup

  1. Start Neo4j locally or in Aura.
  2. Load the schema in graph/neo4j_schema.cypher.
  3. Use graph/sample_queries.cypher to explore bridge creators, awards, and genre clusters.
  4. Expand the demo CSVs with verified public data if you want a larger production-ready graph.

Project structure

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

Sample questions the system can answer

  • 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?

Data ethics and limitations

  • 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.

Portfolio relevance

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

Hybrid web experience

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.py runs the deeper Streamlit strategy workspace with fit scoring, bridge analysis, and graph explanations.

Suggested deployment setup

  • 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.py as the analyst console

About

Graph-based media intelligence prototype for exploring creative networks, cross-market talent bridges, and content strategy across Chinese and American film industries.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors