Investigating the Vocabulary Hierarchy of Odeuropa Explorer

Introduction

Odeuropa Explorer is a fascinating project that digitizes Europe’s olfactory heritage. Funded by the EU’s Horizon 2020 research program, it provides a platform for cross-searching and exploring historical smell experiences.

This project classifies smell-related information into three main categories:

• Smell sources: Objects and substances that emit odors
• Fragrant Spaces: Places and spaces associated with smells
• Gestures and Allegories: Gestures and allegorical expressions related to smell

This article reports the results of investigating the hierarchical structure of these vocabularies using SKOS (Simple Knowledge Organization System) format data published in the Odeuropa vocabularies repository.

Investigation Method

SKOS Hierarchy Visualization Script

To understand the hierarchical structure of the vocabularies, I created a Node.js script to parse SKOS Turtle files.

Key processing steps of this script:

1. Loading the RDF graph: Uses rdflib to parse vocabulary files in Turtle format
2. Extracting hierarchical relationships: Manages skos:broader and skos:narrower relationships with Map objects
3. Identifying top-level concepts: Detects concepts without skos:broader
4. Recursive hierarchy display: Visually represents the tree structure

Execution Method

The script was executed for all three vocabularies:

Investigation Results

1. Smell Sources

Distribution of hierarchy depth:

• Level 0 (top-level): 13 concepts
• Level 1: 383 concepts
• Level 2: 202 concepts
• Level 3: 42 concepts
• Level 4: 45 concepts

Top-level categories:

Examples of characteristic deep hierarchical structures:

Points of interest:

1. Deepest hierarchical structure: Maximum depth of 4, making it the most complex hierarchy among the three vocabularies
2. 13 top-level categories: Systematic classification into Abstract, Artefact, Being, Body, Element, Flora, Food, Fragrance/Cosmetic, Fumes, Matter, Nature, Product, and Religion
3. Overwhelming scale: 685 concepts far exceeds the other two vocabularies (138 and 36)
4. Taxonomic approach: Being and Flora have detailed hierarchical structures close to scientific classification
5. Fusion of culture and science: Covers a wide range of knowledge domains from religious concepts to chemical substances
6. Detailed classification of artefacts: The Artefact category includes diverse artefacts related to smell, such as perfume bottles, jewelry, and smoking implements

2. Fragrant Spaces

Distribution of hierarchy depth:

• Level 0 (top-level): 91 concepts
• Level 1: 44 concepts
• Level 2: 3 concepts

Full hierarchical structure:

Highlights of characteristic hierarchical structures:

Points of interest:

1. Multilingual nature: German (Fabrik, Garten, Wald) and English are mixed, reflecting the characteristics of a European multilingual project
2. Relatively flat structure: Maximum depth of 2 is shallow, with most concepts at the top level or one level below
3. Diversity of places: Comprehensively covers places related to smell, including factories, markets, religious facilities, and natural environments
4. Detailed classification of industrial facilities: Factories (Fabrik) in particular have 7 specialized subcategories, suggesting the importance of smell during the Industrial Revolution

3. Gestures and Allegories

Distribution of hierarchy depth:

• Level 0 (top-level): 34 concepts
• Level 1: 2 concepts

Full hierarchical structure:

Concepts with hierarchy:

1

Major concepts (by category):

• Nose-related gestures: Holding one’s nose, Hand towards the nose
• Hygiene and cleaning activities: Garbage collecting, Street sweeping, Washing
• Medical activities: Examination of urine, Embalming, Doctor sniffing cane
• Religious rituals: Burnt offering, Per fumum (through smoke)
• Everyday activities: Eating, Smoking, Defecation

Points of interest:

1. Nearly flat structure: 34 out of 36 concepts are top-level, indicating low need for hierarchical organization
2. Body action-centric: Many relate to physical gestures and actions
3. Medicine and hygiene: Includes historical medical and hygiene practices such as urine examination and fumigation disinfection
4. Cultural diversity: Covers a wide range of cultural contexts from religious rituals to everyday activities

Technical Considerations

Comparison Between Vocabularies

Comparing the three vocabularies clearly reveals the characteristics of each domain:

Significance of scale differences:

• Smell sources being overwhelmingly large indicates the great diversity of smell sources
• Fragrant Spaces is medium-sized, dealing with concrete concepts of places
• Gestures and Allegories is the smallest but carefully selects culturally important actions

Significance of hierarchy depth:

• Deep hierarchy (Smell sources): Knowledge that is inherently hierarchical, such as natural classification (taxonomy) and product categories
• Shallow hierarchy (Fragrant Spaces, Gestures): Places and actions exist as relatively independent concepts

Alignment with SKOS Design Principles

The three vocabularies investigated have structures faithful to SKOS design principles:

1. Appropriate hierarchy depth: Maximum depths of 1-4, adopting depths suited to each domain’s nature
2. Clarity of broader/narrower relationships: The hierarchical relationships of each concept are clearly defined
3. Balance of top-level concepts: Smell sources has 13 systematic categories, Fragrant Spaces has 91 diverse places, Gestures has 34 independent actions

Data Modeling Characteristics

The differences in hierarchy depth across the three vocabularies reflect the nature of each domain:

• Smell sources (depth 4): Knowledge systems that are inherently hierarchical, such as natural classification and product hierarchies
  • Example: Being -> Tier -> Wirbeltiere -> Mammal -> Lion
  • Example: Flora -> Baume -> Holz -> Sandelholz
• Fragrant Spaces (depth 2): Shallow hierarchy based on place containment relationships
  • Example: Building -> House -> Apartment
• Gestures and Allegories (depth 1): Exist as independent actions with low need for hierarchical organization
  • Example: Smelling -> Piss smelling (one of the few hierarchies)

Implementation Considerations

Key points in script implementation:

1. Circular reference checking: Uses a visited Set to prevent infinite loops
2. Label retrieval: Fallback from URI when skos:prefLabel is not available
3. Sorting: Alphabetical sorting ensures output consistency
4. Statistical information: Depth distribution and child element count aggregation enable quantitative understanding of vocabulary characteristics

Summary

Through investigation of the three Odeuropa Explorer vocabularies (Smell sources, Fragrant Spaces, Gestures and Allegories), the following was found:

1. Design tailored to each domain:

   • Smell sources (685 concepts, depth 4): Systematic classification and deep hierarchy
   • Fragrant Spaces (138 concepts, depth 2): Moderate hierarchy and diversity
   • Gestures and Allegories (36 concepts, depth 1): Flat structure and independence

2. Cultural and historical richness: The diversity of European olfactory cultural heritage is condensed into a total of 859 concepts

3. Multilingual nature: German, English, French, and Latin are mixed, embodying the characteristics of a pan-European project

4. Flexibility of hierarchical structure: Optimal hierarchy depths are adopted for each domain, from natural classification to places and actions

5. Balance of practicality and scholarship: A balance between flat structures suitable for search and exploration and hierarchy necessary for knowledge systematization

Understanding such vocabulary hierarchical structures allows us to learn practical examples of knowledge organization in digital humanities. The analysis of three vocabularies with different characteristics demonstrates that SKOS is an effective tool not only for academic research but also for cultural heritage digitization projects.

Reference Links

• Odeuropa Explorer - Olfactory heritage exploration platform
• Odeuropa Vocabularies GitHub - Repository analyzed in this investigation
• SKOS Simple Knowledge Organization System - W3C SKOS specification
• Odeuropa Project - Official project site

Investigation Data and Code

Files used and generated in this investigation:

Analysis Targets

• Odeuropa vocabularies - SKOS Turtle format vocabulary data

Generated Files

• visualize-hierarchy.js - SKOS hierarchy visualization script
• olfactory-objects-hierarchy.txt - Complete hierarchy output for Smell sources (685 concepts)
• hierarchy-fragrant-spaces-20251010-231358.txt - Complete hierarchy output for Fragrant Spaces (138 concepts)
• olfactory-gestures-hierarchy.txt - Complete hierarchy output for Gestures and Allegories (36 concepts)