Smell Explorer

The Odeuropa Smell Explorer [https://explorer.odeuropa.eu] is a brand-new web tool developed for the exploration of smell as a cultural phenomenon. This searchable website enables you to discover the smells the past and understand how they shaped European history and culture. The Smell Explorer is the result of three years of intensive research and development by an international team of computer scientists, AI experts and humanities scholars. Its target audiences are scholars, perfumers, heritage professionals, artists, and basically anyone with an interest in the world of scents, in olfactory language and imagery, and in the important role scents play in our daily lives.

Where can I find the Explorer?

The Smell Explorer is accessible online at https://explorer.odeuropa.eu. The underlying open access data and open source software can be found on Odeuropa’s Github repositories.

How was the Explorer developed?

The data in the Smell Explorer was extracted from 23,000 images and 62,000 historic texts in six languages (English, Italian, French, Dutch, German and Slovene) available in the public domain.

The Odeuropa computer vision team developed techniques to detect objects associated with smell in images. After a set of 5000 images were manually annotated, the team trained the computer to recognise similar smell-related elements and used machine learning techniques to expend the database. Thus, the computer captures ‘smell sources’ (odorants and objects with smells such as perfume bottles), fragrant places and smell gestures (smelling, pinching the nose).

To complement the images with mentions to smells and smell experiences in texts, our computer science teams manually annotated thousands of historic books including novels, theatre scripts, travel writing, botanical textbooks, court records, sanitary reports, sermons, and medical handbooks. Using these examples, we developed an automated system that can replicate the manual annotation by identifying in texts smell-related information (who smells what where and what characteristics are used to describe the scent?), including smell-related emotions.

Our semantic web experts have used these results to develop a user-friendly interface to search and browse through smell-related text and images. For this, they first designed an event-based ontology. The Explorer is organised as a Knowledge Graph, a semantically rich database where the multi-modal information has been curated and stored, following semantic web standards. In July 2023 it already held nearly 800.000 olfactory depictions and descriptions.

What kind of questions can the Smell Explorer answer?

What are the most significant smells and smellscapes of Europe? What scents are most discussed in historical texts and what reactions did they evoke? What odorants were used in a pomander or in potpourri? What olfactory objects and fragrant places are most depicted in paintings and prints? Are woody smells more appreciated than fruity, or animalistic smells? How did ‘nose witnesses’ describe their smell experiences? Are smell words in Italian comparable to those in English, French, Slovenian or Dutch? Where could one smell civet or sulphur in the 18th century? What emotions can be evoked by the smell of incense? How are scents classified?

The Odeuropa Smell Explorer can provide answers to all these questions. It is the first database that can be queried ‘nose-first’ (using the sense of smell as an entry point), thanks to the unique technologies used to design the tool. This makes it a valuable resource for anyone interested in understanding how the past smelled and how smell experiences were described and depicted.

Whereas most websites focus on fine fragrances, the Explorer will help you nosedive into all kinds of significant smellscapes, foul or fragrant. It can help perfumers discover how scents were classified into different scent families and what sentiments were attached to what aromas. It can help scientists compute odour landscapes. Art historians may find new olfactory allegories and iconographies. Aromatherapists can discover ancient recipes. The Explorer generates insightful quotes about the values of smell. In general, the Explorer provides a unique opportunity to tap into the nose wisdom of the past. You can also review the competency questions the Odeuropa team used to develop the Explorer.

How can I browse or search the database?

The Smell Explorer provides two major options for querying the data:

1. Browse by smell sources and fragrant spaces

The Explorer provides the opportunity to browse over 550 smell sources, 115 fragrant spaces and 35 olfactory gestures. Separate pages give entry to these searches. The Smell Source page lists objects and substances which are recognised to emit on odour. Visitors can order them alphabetically or by occurrence. Clicking on a smell source will lead to a new page dedicated to that source. This provides: 1. A timeline of the descriptions of the source, 2. A map locating the references, 3. A word cloud summarizing the most used characteristics to describe the source, 4. An overview of all the textual and visual occurrences of the source. The ‘Fragrant Spaces’ (smellscapes) page follows the same rationale. A page with an overview of olfactory gestures will be added later.

2. Use the elaborate search environment

In the open search environment visitors can search the explorer for either texts or images, or both. You can search for the word ‘incense’ (which will mostly deliver English language results), or for the category ‘rose’ (which also provides results for the other 6 languages incorporated in the Explorer). Searches can be further limited by categories such as language, collection, place (where the smell is mentioned), time (year, period or season), or emotion.

The Explorer displays your findings as cards or text snippits. Each ‘nose witness report’ (be it textual or visual) is presented in its original context: the book from whence the quote comes, or the painting that depicts the smell source. Permalinks provide the opportunity to browse to the original digital heritage collection which holds the resource.

Can I download the results or design more elaborate queries?

Yes you can! You can sign up as a member. This creates the opportunity to save your results into your own smell library, and export them. Furthermore, the European Olfactory Knowledge Graph, which forms the ‘data backbone’ of the Explorer can be queried through a SPARQL endpoint. The overview of the vocabularies and taxonomies we incorporated in the Knowledge Graph may provide a starting point for queries of a more flavour analytics nature. By engaging with these datasets you can pose queries such as ‘What descriptions of smell qualities can be found to coincide with the Flavornet database ‘Odor’ column or with Dravnieks 146 odour descriptors?’ ‘Is the smell of goats more often described through the Linaeus taxonomy as caprylic, or through the Dravnieks descriptors as rancid?’

Can you help me calculate the data?

Yes we can! The Odeuropa team has developed interactive nosebooks which can make odour analytics easier for you. These nosebooks can for instance provide a spreadsheet with all the relevant quotes about a specific smell source, listing: Smell Word, Smell Source, Quality, Location, Perceiver, Time, Effect, SentenceBefore, Sentence, SentenceAfter, Year. You will also able able to review n-Gram visualisations of the relative frequency of their use over time. Furthermore, you can extracts smells from your own text or images. More information can be found at https://odeuropa.eu/nosebooks/.