Ontology and Knowledge Graph of 22 Favorite English Songs

This project builds an RDF knowledge graph of 22 favorite English songs using semantic web technologies, combining external vocabularies (e.g., schema, dbr) with custom ontologies (myOnto) and personal resources (mySemantics). The graph is designed, queried, and visualized through SPARQL, SHACL, and GraphDB tools.

Course: Semantic Web and Linked Data
Institution: University of Zurich (Spring 2023)
Contributors: Xiaojing Zhang and Huan Liu

---

Project Structure

project-root/
│
├── classes_properties.ttl      # Custom ontology (myOnto) definitions and property constraints
├── resources.ttl               # RDF triples representing 22 songs and related entities
├── metadata.xlsx               # Song metadata collected from Wikipedia and Spotify
├── data_model.graphml          # Ontology diagram for visualization in GraphDB or Gephi
│
├── script.ipynb                # Notebook for uploading RDF data and running SPARQL queries
├── shacl.ipynb                 # SHACL validation rules and conformance report
│
├── Project Report.pdf          # Full report describing data modeling, queries, visualizations
└── README.md                   # This file

---

Project Goals

• Represent 22 English songs as RDF resources using Turtle format
• Define and apply a custom ontology (myOnto) for music metadata
• Integrate resources from DBpedia (dbr) and schema.org with manually created ones (mySemantics)
• Query the graph using SPARQL and SPARQL*
• Validate structure and constraints using SHACL
• Visualize structure and genre distributions in GraphDB

---

Ontology: myOnto

• Custom classes: myOnto:Song, myOnto:Person, etc.
• Custom properties: myOnto:hasName, myOnto:hasWikipageID
• Constraints:
  • myOnto:Person must have exactly one hasWikipageID (type: integer)
  • Uses owl:FunctionalProperty and owl:InverseFunctionalProperty

---

Data Sources

• Wikipedia
• Spotify
• MusicBrainz / Discogs (selective)
• 22 song entries with full metadata:
  • Title, release date, artist, writer, producer, genre, label, etc.

---

SPARQL Queries (in Project Report)

Includes:

• Top 3 most recent songs
• Songs released after 2017
• Songs by Ed Sheeran
• Songs of genre Pop
• A manually constructed song via SPARQL INSERT
• A subgraph extraction using CONSTRUCT
• Two queries using SPARQL*

---

SHACL Validation

• Implemented in shacl.ipynb
• Checks:
  • Cardinality constraints
  • Datatype restrictions
  • Class membership

---

Visualizations

• Created in GraphDB and Gephi from data_model.graphml
• Highlights:
  • Class centrality: Songs are the most connected nodes
  • Genre distribution: 11 genres, Pop is most represented
  • SPARQL CONSTRUCT-based subgraphs (e.g., only recent Pop songs)

---

RDF* / SPARQL* Examples

:Xiaojing :listensTo << dbr:Easy_on_Me dbo:releaseDate "2021-10-15"^^xsd:date >> .

Captures listener behavior as metadata about a triple.

Takeaways

• Semantic technologies can represent music metadata flexibly and precisely
• Custom ontologies complement external vocabularies for personalization
• SPARQL* and SHACL enrich graph querying and validation
• RDF graphs provide a strong foundation for knowledge-based music apps

Tools Used

• RDF & Turtle via Protégé and text editors
• GraphDB for graph storage and visualization
• Jupyter for SHACL and SPARQL execution
• Gephi for .graphml ontology visualization

License

• This repository is for educational and academic use only.
• No copyrighted data.

How to Reuse

To explore or reuse this project:

• Clone the repository
• Open script.ipynb or shacl.ipynb in Jupyter
• Visualize data_model.graphml in Gephi or GraphDB
• Modify resources.ttl and classes_properties.ttl to extend the ontology