Odeuropa: The World of Linked Data for Extracting Scents from Historical Documents

Introduction

Odeuropa is a unique project that extracts descriptions of “scent” from European historical documents and structures them as Linked Data. In this article, we explore the actual data through the SPARQL endpoint, revealing its structure and design philosophy.

What is Odeuropa?

• Project name: Odeuropa (Odeurs d’Europe = European scents)
• Database URL: https://data.odeuropa.eu/
• SPARQL endpoint: https://data.odeuropa.eu/repositories/odeuropa
• Web interface: https://explorer.odeuropa.eu/

Overall Data Model

Odeuropa uses an extended ontology specialized for scent, based on CIDOC-CRM (Conceptual Reference Model for Cultural Heritage).

Key Concepts and Relationships

Key points:

• Fragment directly references Emission, Smell, and Experience
• Object is accessed via Emission (Fragment -> Emission -> Object)
• Emission plays the central role of causally connecting Object and Smell

Learning Data Structure Through Examples

Let’s examine the data structure using the 1810 German agricultural book “Grundsatze der rationellen Landwirthschaft” (Principles of Rational Agriculture) as an example.

1. Source (Document)

An entity that stores basic information about the document.

Run query

Key properties:

• rdfs:label: Title
• schema:author: Author (Albrecht Daniel Thaer)
• schema:dateCreated: Creation year (1810)
• schema:inLanguage: Language (de)
• schema:genre: Genre (Household texts & recipes)
• schema:locationCreated: Place of creation
• P106_is_composed_of: Contained fragments

2. Fragment (Text fragment)

A portion of text containing scent-related descriptions.

View data

Meaning of the text: “They (clays) become slippery and more elastic when moistened, emit a clay smell, and dry into solid but more crumbly clumps.”

Key properties:

• rdf:value: Actual text content
• schema:position: Position within the document (4th fragment)
• P106_is_composed_of: Important words contained (“Sie”, “Thongeruch”)
• P67_refers_to: Referenced concepts (Emission, Smell, Experience)
• P165i_is_incorporated_in: Parent document (Source)

3. Emission (Scent emission event)

Represents an event where a scent is generated. Emission is referenced from Fragment and connects Object and Smell.

View data

Key properties:

• F3_had_source: Source of the scent (Object “Sie”)
• F1_generated: Generated smell (Smell “Thongeruch”)
• P92_brought_into_existence: Smell brought into existence
• P12_occurred_in_the_presence_of: Things present at the event (Object and Smell)
• time:hasTime: Time of occurrence (1810)
• P67i_is_referred_to_by: Fragment referencing this Emission

Role of Emission: Emission is the central event expressing the causal relationship of “which Object (source) generated which Smell (scent), and when.”

4. Object (Scent source)

The object or substance that emits the scent. Object is referenced from Emission.

In this example, “Sie” (they) is a pronoun referring to clay or soil.

Types:

• S10_Material_Substantial: Material substance
• S15_Observable_Entity: Observable entity

Key properties:

• rdfs:label: Object name (“Sie”)
• P12i_was_present_at: Event where this object was present (Emission)

Connection path:

View data

5. Smell (Scent)

The central concept representing the scent itself.

Key properties:

• rdfs:label: Name of the scent (Thongeruch = clay smell)
• P92i_was_brought_into_existence_by: Emission that generated this smell
• P140i_was_attributed_by: Experience that recognized this smell

View data

6. Experience (Scent experience event)

An event where a person perceives or experiences a scent.

Key properties:

• F2_perceived: Perceived smell
• O8_observed: Observed smell
• P140_assigned_attribute_to: Target to which attributes were assigned
• P14_carried_out_by: Experiencer (Actor)

View data

Data Flow: The Complete Story

Data flow explanation:

1. Fragment directly references three concepts (Emission, Smell, Experience)
2. Emission is the center of causal relationships:
   • From Object (source)
   • Generates Smell (scent)
3. Experience perceives Smell
4. Object is indirectly connected to Fragment via Emission

SPARQL Query Examples

Searching by Language

When searching by German label:

Retrieving Visual Items with Images

Avoiding duplicates when multiple images exist:

0

Retrieving Smells and Their Sources

1

Ontologies Used

CIDOC-CRM

• E33_Linguistic_Object: Linguistic object (document)
• E36_Visual_Item: Visual item
• E39_Actor: Person (author, observer)
• E53_Place: Place
• E77_Persistent_Item: Persistent item
• P67_refers_to: Refers to
• P106_is_composed_of: Is composed of
• P140_assigned_attribute_to: Assigned attribute to

CRMsci (Scientific Observation Extension)

• S10_Material_Substantial: Material substance
• S15_Observable_Entity: Observable entity
• O8_observed: Observed

Odeuropa Custom Extensions

• L12_Smell_Emission: Smell emission
• F1_generated: Generated
• F2_perceived: Perceived
• F3_had_source: Had source

Schema.org

• schema:author: Author
• schema:dateCreated: Creation date
• schema:inLanguage: Language
• schema:genre: Genre
• schema:image: Image
• schema:position: Position

Significance of the Project

The Odeuropa project is groundbreaking in the following ways:

1. Digitization of sensory data: Structures “scent,” sensory information that was previously difficult to digitize
2. Application to historical research: Enables analysis of what past people perceived as scent and how
3. Linked Data in practice: Implementation of advanced Semantic Web technology using CIDOC-CRM
4. Interdisciplinary approach: Fusion of history, information science, and sensory studies

Summary

The Odeuropa database is an ambitious project that uses text mining, ontology design, and Linked Data technologies to extract and structure the abstract concept of “scent” from historical documents.

While based on the established cultural heritage ontology CIDOC-CRM, it achieves a reusable and extensible data model by adding scent-specific concepts (Emission, Experience).

This approach can be applied to the digitization of other sensory information (sound, taste, touch, etc.), demonstrating new possibilities for digital humanities.

References

• Odeuropa Explorer: https://explorer.odeuropa.eu/
• SPARQL Endpoint: https://data.odeuropa.eu/repositories/odeuropa
• CIDOC-CRM: http://www.cidoc-crm.org/
• Document used as example: https://www.deutschestextarchiv.de/book/show/thaer_landwirthschaft02_1810