Knowledge Graph engineering

Tourism and Waste Management

🧭 Project Objective

This project builds a knowledge graph (KG) that provides insights into the relationship between tourism and waste management in the Province of Trento, Italy. The KG supports stakeholders such as tourists, facility owners, policymakers, and researchers by offering:

Locations of waste bins and recycling facilities.
Waste types and disposal methods.
Analysis of waste production trends, particularly in tourist areas.
Evaluation of waste-related policies and practices

⚙️ Methodology: The iTelos Phases

Project Architecture

The project is developed using iTelos Methodology

iTelos is a methodology designed to facilitate the reuse of existing knowledge—ontologies, schemas, and datasets—by guiding the development of Knowledge Graphs (KGs) through a purpose-driven, middle-out approach.
Its aim is to make both the developed applications and their knowledge components reusable and shareable.
It consists of 5 phases: Purpose definition, Information Gathering, Language Definition, Knowledge Definition and Entity Definition.

Methodology

Purpose Definition

Defined the project’s informal goal: exploring the relationship between tourism and waste management in Trentino.
Identified user scenarios and personas (e.g., tourists, policymakers, facility owners).
- Six personas were selected representing stakeholders (tourist, facility owner, researcher, analyst, etc.)
- Realistic use cases listed (e.g., ski tourists needing recycling info, policy-makers analyzing pandemic waste trends).
Formulated Competency Questions (CQs) to drive the design of the Knowledge Graph.

Information Gathering

Collected, cleaned, and standardized data from various open sources:
- OpenStreetMap (waste bins, tourist sites)
- Dolomiti Ambiente (waste categories)
- ISTAT (population data)
- ISPRA (waste production statistics)
Created intermediate datasets and merged them based on shared identifiers like municipality codes.

Language Definition

Defined formal concepts, properties, and data values using:
- UKC (Universal Knowledge Core) for generic terms.
- OpenStreetMap tags for domain-specific terminology.
Introduced custom definitions where existing vocabularies were insufficient (e.g., geometry, waste disposal types).

Knowledge Definition

Modeled the teleology (bottom-up) and teleontology (top-down) of the domain.
Combined both into a final teleontology using a middle-out approach.
Used Protégé for ontology modeling and created schema alignment with reference ontologies like schema.org.

Entity Definition

Linked real-world data to the ontology using the Karma tool.
Ensured each entity had unique identifiers and consistent relationships.
Exported the final Knowledge Graph in RDF Turtle (TTL) format.

Project Details

🧑‍🤝‍🧑 Personas & Scenarios

The project was grounded in six personas and six real-world scenarios, including:

Tourists needing to locate recycling bins (e.g., Maximilian at Monte Bondone).
Facility owners complying with waste regulations (e.g., Luciana).
Policymakers analyzing tourism’s impact on waste (e.g., Diego).
Researchers studying COVID-19 effects on waste generation (e.g., Chiara).

❓ Competency Questions (CQs)

Examples of key questions the KG can answer:

Where are the nearest recycling bins for glass/plastic?
What types of waste are most common in ski resorts?
How did COVID-19 impact waste generation?
What special waste disposal facilities exist in Trentino?

🗺️ Data Sources

OpenStreetMap (OSM): For geospatial data on bins, facilities, and tourist attractions.
Dolomiti Ambiente: For waste type and disposal method guidelines.
ISPRA: For annual waste production statistics.
ISTAT: For municipality population data.
Overpass Turbo: For querying OSM and extracting detailed points of interest.

🧠 Knowledge Graph Design

Entities: Person, Municipality, Location, TouristAttraction, Waste, WasteBasket, WasteProduction, WasteDisposalType.
Properties: Names, coordinates, categories, population size, disposal methods, etc.
Teleontology: Modeled with Protégé, extended schema.org and domain-specific classes.
SPARQL Queries: Executed to verify the KG's ability to answer competency questions.

📈 Results & Evaluation

Successfully integrated and cleaned multiple open datasets.
Designed and implemented a rich knowledge graph schema.
Evaluated for purpose coverage (~70%) and full ontology reusability (100% for referenced ontology classes).
Good connectivity (Entity Connectivity Score = 9.5).

🚧 Challenges & Open Issues

OSM data sparsity for some bin types.
Lack of standardized waste policy data across municipalities.
Potential improvements include:
- Integrating real-time data (e.g., smart bin sensors).
- Extending coverage to additional regions.
- Developing an API or web tool for live querying.