How Multi-Modal Knowledge Graphs Are Transforming Sports Media

Using a Multi‑Modal Content Graph and AI across your sports media delivers the ultimate sports knowledge hub – unlocking massive latent value in video, audio, imagery, statistics, and editorial content.
Modern sports organizations sit on vast archives of matches, highlights, interviews, commentary, statistics, and written analysis. Yet, most of this content remains siloed, under‑described, and difficult to explore beyond simple keyword search. This is where Data Graphs – an AI‑powered, enterprise‑grade, multi‑modal knowledge graph – fundamentally changes what is possible.
This article expands on the demo outlined below and shows how Data Graphs can be used to explore, analyze, and increase the value of sports media by combining video, sports data, and AI into a single, connected knowledge fabric.

Game Strategy

The primary mechanism for unlocking value is to join and align sports media with sports data metadata using a structured domain model (ontology) for both media and sport.
For this work, we use the IPTC Sports Schema, a well‑designed, open, and extensible model for representing:

• Sports competitions and seasons
• Matches and events
• Teams, athletes, and officials
• In‑game actions (goals, penalties, substitutions, cards, etc.)

In Data Graphs, the IPTC Sports Schema is combined with:

• CreativeWorks and Media Ontologies (video, audio, images, articles)
• Temporal and spatial concepts (time‑coded events, positions on the pitch)
• AI‑generated annotations and embeddings

With respect to sports data enrichment, there are two complementary strategies:

1. Using a sports data provider to augment video and media with structured match and event data.
2. Using AI vision and media analysis services (for example, our partner’s Bitmovin AI) to analyze video directly and automatically generate metadata.

The result is a single, queryable, explainable graph that connects what happened, who was involved, and where it appears in the media.

Warm Up – The Demo

Using 2015–2017 Premier League and Champions League video and data from SoccerNet, we created a practical demonstration of how Data Graphs can deliver the ultimate AI‑powered sports knowledge hub.
This demo combines:

• Hundreds of full match videos
• Structured match, team, player, and event data
• The IPTC Sports Schema
• Data Graphs’ hybrid GraphRAG AI capabilities

Data Graphs is much more than a graph database. Its native handling of video, audio, text, and embeddings makes it a true content graph for:

• Analyzing and querying media
• Managing and enriching metadata
• Streamlining content production workflows
• Powering new AI‑driven applications and fan experiences

Users can ask natural‑language questions such as:

• “Show me all goals scored by Liverpool in the Champions League with the video clips.”
• “Compare Ronaldo and Messi’s performances with key moments.”
• “Find all red cards in away matches and the associated footage.”

Behind the scenes, these questions are answered by combining semantic vector search with precise graph queries over structured sports data.

First Half – Building the Sports Content Graph

Domain Model: IPTC Sports Schema + Media Ontologies

At the core of the system is a clean, explicit domain model. The IPTC Sports Schema provides the backbone for sporting concepts, while CreativeWorks and Video ontologies describe the media itself.
This allows Data Graphs to natively represent relationships such as:

• A Match has many Events
• An Event involves Athletes and Teams
• An Event is linked to a precise Video Segment
• A Video is part of a Competition and Season

Because the model is explicit, it remains explainable, extensible, and portable across competitions and sports.

SoccerNet Video and Sports Statistics

For the demo, we used:

• SoccerNet match videos
• SportsMonks statistics and event data

SportsMonks aligns closely with the IPTC model, making ingestion straightforward – but any sports data provider can be used. Data Graphs is intentionally schema‑flexible while still enforcing semantic consistency.

Generating Video Moments from In‑Game Actions

A critical step is linking structured events (goals, fouls, substitutions) to time‑coded moments in video.
Each in‑game action becomes a first‑class node in the graph, connected to:

• The match timeline
• The athletes and teams involved
• One or more video segments

This instantly turns raw match footage into a searchable, explorable library of meaningful moments.

Querying the Graph: GQL and OpenCypher

Because everything lives in a graph, powerful queries become simple and expressive:

• Find all goals by a specific player across competitions
• Compare team performance home vs away
• Retrieve all penalties conceded in the last 10 minutes of matches

Data Graphs supports GQL/OpenCypher, making it easy for developers, analysts, and AI agents to programmatically explore the data.

Visual Graph Exploration

Beyond code, Data Graphs provides visual graph exploration, allowing users to:

• Navigate from a match to events, players, and clips
• Discover unexpected connections
• Validate data quality and completeness

This is invaluable for editors, analysts, and archivists who want insight without writing queries.

Second Half – AI, GraphRAG, and Contextual Views

Dynamic Agentic MCP‑Based AI Configuration

Data Graphs includes a Model Context Protocol (MCP)‑based AI architecture that enables multiple specialized agents to operate on the same graph.
Agents can be configured to:

• Answer fan‑facing questions
• Support editorial research
• Assist performance analysts
• Drive internal content operations 

True Hybrid GraphRAG

Unlike vector‑only RAG systems, Data Graphs delivers true hybrid GraphRAG out of the box:

• Semantic retrieval over vectorized graph content
• Precise structured retrieval using Cypher/GQL
• Parameterised, reusable MCP tools for domain‑specific queries

This means an AI assistant can seamlessly combine statistical accuracy with rich contextual understanding. The Data Graphs native AI assistant operates using this architecture.

Contextual Rendered Views

One of the most powerful features is the ability to render views contextually. Instead of returning raw JSON or text, Data Graphs can render rich, contextual views for:

• Matches and events
• Teams and athletes
• Competitions and seasons

These views dynamically pull together data, media, and AI insights from across the graph – creating human‑friendly representations that can be embedded into applications, dashboards, or editorial tools.
In the example rendered view shown above, a single match concept (English Premier League 2015–2016, Chelsea v Aston Villa) becomes the organising context for:

• Match metadata (competition, season, date, score)
• Linked full‑match videos (first half and second half)
• A structured match timeline of events, such as goals, assists, substitutions, and cards
• Direct navigation into underlying concepts (players, teams, events)
• An interactive graph view showing how events, athletes, teams, and actions are connected

Each timeline entry is not just a label, but a graph entity linked to precise video moments, participants, and statistics. Users can move seamlessly from a goal in the timeline to the video clip, to the player profile, and out into the wider competition graph.
This approach replaces hard‑coded pages with schema‑driven, graph‑powered views that adapt automatically as new data, competitions, or media are added.

Automated Metadata with our partner Bitmovin and the Bitmovin Content Analysis AI 

AI video analysis services such as Bitmovin's Content Analysis AI can be integrated to automatically:

• Detect key in‑game actions
• Recognize players and teams
• Identify highlight moments

This metadata flows directly into the graph, continuously enriching the knowledge base and reducing manual annotation costs.

Cup Winners – Where the Value Is Unlocked

A multi‑modal sports knowledge graph enables:

• Better content analysis and faster editorial workflows
• Sports performance analysis augmented with video evidence
• Richer fan engagement through interactive, AI‑driven experiences
• Compelling storytelling with instant access to historical context
• Rapid research for pre‑ and post‑match analysis
• Brand and sponsor activation through targeted, contextual media
• Long‑term sports archives that are searchable, explorable, and monetizable

In short, it turns passive sports media archives into active, intelligent assets.

Other Sports

The IPTC Sports Schema is sport‑agnostic and competition‑agnostic. The same approach applies to:

• Football, basketball, rugby, cricket, tennis
• Individual and team sports
• Leagues, tournaments, and international competitions

Simply feed Data Graphs with your media and sports data, and the platform does the rest – connecting, enriching, and activating your content with AI.

Data Graphs shows how multi‑modal knowledge graphs are no longer experimental technology – they are a practical foundation for the next generation of sports media, analytics, and fan experiences.

See how it works - get in touch for a demo!