Overview

I had the opportunity to try tools related to Odeuropa, so this is a memo about that experience.

What is Odeuropa?

An explanation is available on the following page.

https://odeuropa.eu/

Below is a machine-translated description.


Odeuropa is an innovative EU-funded project that studies Europe’s “olfactory cultural heritage.”

Project objectives: To investigate and document the role that smells have played in European culture from 1600 to 1920. Using cutting-edge AI technology, it extracts smell-related information from approximately 43,000 images and 167,000 historical texts (in English, Italian, French, Dutch, German, and Slovenian).

Major deliverables:

  1. Smell Exploration Engine - A unique website that allows searching through over 300 years of European olfactory history

  2. Encyclopaedia of Smell History and Heritage - An online reference work on smell-related cultural phenomena written by experts

  3. Olfactory Storytelling Toolkit - A practical guide for museums and cultural heritage institutions to incorporate smells into exhibitions

  4. Heritage Smell Library - A collection of historically significant smells recreated in collaboration with perfumers and preserved

This project enables richer, multi-sensory historical experiences by adding the new dimension of “olfaction” to the traditionally visual and auditory understanding of cultural heritage.

Tools

Several related tools are introduced on the following page.

https://odeuropa.eu/nosebooks/

Live Image Processing Demo

I used the following image published on the National Diet Library’s “Portraits of Modern Japanese” page.

https://www.ndl.go.jp/portrait/datas/224

Access the following and specify the image URL.

https://huggingface.co/spaces/mathiaszinnen/odeuropa-demo

https://www.ndl.go.jp/portrait/img/portrait/0224_6.jpg

The result with “Confidence Threshold” set to 0.2 is shown below. It was able to recognize the cigarette held in the hand.

Smells Extraction

This appears to be a tool for extracting smells from text.

https://smell-extractor.tools.eurecom.fr/

I had AI create a sample English text containing descriptions of smells.

ATshiIswmaolrkneidngpassttarttheedbdaikfefreyr,entthleywtahromugahr.omTaheofrifcrhe,shnluyttbyakaerdombareoafdfarnedshcliynngarmoounndroclolfsfeweafbteeadnsthpreorumgehattehdetahier,kimtackhienng,miytsmobuitthtewra-tsewreeitnsstcaennttlyp.roLmaitseirngthtahteapfetrefrencotons,tawrhtentoImsytedpapye.dIotutwsaisdesuacfhtearctohnetrsausmtmetrosyteosrtme,rdtahyeepveetnriincghowrhefnilIlehdadmywannodsetrreidlsth-rotuhgahttdhiestgianrcdteinv,ewehaerrtehythsecehnetadoyfprearifnumoenodfryjassomiilnemibxlionsgsowmisthhutnhgehsewaeveytifnratghreanacier,ofinwteetrmgirnagslsi.ngItwirtehmitnhdeedshmaerpo,fgmryeevnissimteltlootfhefroelsdhllyibcruatrygrlaassstawnedekt,hewhseurbetlteheswmeuesttnyesssmeolflroofseasg.edNoptahpienrgacnodulldeahtahveer-pbroeupnadrebdoomkesfhoardtgordeaeyt'esdtmreipastoItehnetesreeadf,ooadccmoamrpkaenti,edhobwyevheirn,tswhoefrevatnhiellpaunagnedntalomdoonrdoffrofmisthheanddecbormipnoesiansgsaluilgtneidnmiynstehneseyselalsowIinngavpiaggaetse.dthroughthecrowdedstalls,thesaltytangoftheoceanmixingwiththemetallicscentoficeandseafood.

The result of registering this text is shown below. Smell-related information such as “Location,” “Quality,” and “Smell Word” was detected.

The results can also be downloaded as Excel, and the following was obtained.

BookSmell_WordSmell_SourceQualityOdour_CarrierEvoked_OdorantLocationPerceiverTimeCircumstancesEffectSentenceBeforeSentenceSentenceAfter
demo0aromaof freshly baked bread and cinnamon rollswarmthrough the airthe bakerymaking my mouth water instantlyAs I walked past the bakery , the warm aroma of freshly baked bread and cinnamon rolls wafted through the air , making my mouth water instantly .Later that afternoon , when I stepped outside after the summer storm , the petrichor filled my nostrils - that distinctive earthy scent of rain on dry soil mixing with the sweet fragrance of wet grass .
demo0scentfragrancepetrichorofof rain on dry soilofof wet grassdistinctive earthysweetofofof
demo0smellhintsof aged paper and leather - bound booksof vanillaalmond from the decomposing ligninmustyto the old librarymelast week
demo0aromascentof freshly ground coffee beansitsrichnuttybittersweetthe kitchen
demo0perfumesmellof jasmine blossomsof freshly cut grasssweetness of rosesheadysharpgreensubtlein the airthe gardenI
demo0odortangscentof fish and brineof the oceanof ice and seafoodpungentsaltymetallicthe crowded stallsmy

Experiment

Setup

I tried to see if a similar approach could be applied to Japanese texts.

The following data from the Koui Genji Monogatari Text Database was used as a sample.

https://kouigenjimonogatari.github.io/

The following text from “34 Wakana Jo, page 1072” was used as the target.

The following prompt was used with Claude Opus 4.

sAmselIlswEaxelcxketelrdacptaisotnthebakery,thewarmaromaoffreshlybakedbreadandcinnamonrollswaftedthroughtheair,makingmymouthwaterinstantly.Laterthatafternoon,whenIsteppedoutsideafterthesummerstorm,thepetrichorfilledmynostrils-thatdistinctiveearthyscentofrainondrysoilmixingwiththesweetfragranceofwetgrass.Itremindedmeofmyvisittotheoldlibrarylastweek,wherethemustysmellofagedpaperandleather-boundbookshadgreetedmeasIentered,accompaniedbyhintsofvanillaandalmondfromthedecomposingligninintheyellowingpages.Thismorningstarteddifferentlythough.Therich,nuttyaromaoffreshlygroundcoffeebeanspermeatedthekitchen,itsbitter-sweetscentpromisingtheperfectstarttomyday.ItwassuchacontrasttoyesterdayeveningwhenIhadwanderedthroughthegarden,wheretheheadyperfumeofjasmineblossomshungheavyintheair,interminglingwiththesharp,greensmelloffreshlycutgrassandthesubtlesweetnessofroses.Nothingcouldhavepreparedmefortoday'striptotheseafoodmarket,however,wherethepungentodoroffishandbrineassaultedmysensesasInavigatedthroughthecrowdedstalls,thesaltytangoftheoceanmixingwiththemetallicscentoficeandseafood.

Results

Below are the smell-related information extracted from the classical text in a similar format:

BookSmell_WordSmell_SourceQualityOdour_CarrierLocationEffectSentenceSentenceAfter
(Unknown)hana (flowers)(Spring flowers in general)chirisukite (having scattered and faded)kosue (treetops), kotachi (grove)hana ha mina chirisukite nakori kasumeru kosue no asamidori naru kotachi mukashi fuchi no e mushi tamahi shi kono koro no koto narikenkashi to oboshiitsurutoshitsuki no tsumorinikeru hodomo sono ori no koto kakitsutsuke ahare ni obosaru
(Unknown)nihohi (fragrance/scent)fuchi (wisteria flowers)woennaranu (indescribably wonderful)kono kake (this shade), kotachi (grove)kokoro sofu (heart is drawn), kono kake wo ha tachi hanarubeki (unable to leave this shade), warinaku itekateni oboshi yasurahi tari (hesitating, unable to depart)kono fuchi yo ikani somekemu iro ni kana woennaranu kokoro sofu nihohi ni koso ikateka kono kake wo ha tachi hanarubeki to warinaku itekateni oboshi yasurahi tariyamakiha yori sashiitsuru hi no hanayaka naru ni sashiahi me mo kakayaku kokochi suru on-sama no koyonaku nehikuhahari tamaeru on-kehahi nato wo mezurashiku hotohetemo mitatematuru ha mashite yo no tsune narasu oboyureha

Supplementary information:

  • Perceiver: From context, the person being seen off by Chunagon no Kimi (likely a woman)
  • Time: Asaborake (dawn), kono koro (spring season)
  • Circumstances: A parting scene at the place where wisteria was planted long ago, as Chunagon no Kimi comes to see the person off
  • Evoked_Odorant: Not mentioned
  • SentenceBefore: Chunagon no Kimi mitatematuri wokuru tote tsuma to oshiaketaru ni tachikaheri tamahite

Discussion

The results would need review by specialists, but they may be usable as preliminary data.

Summary

I tried Odeuropa-related tools and also explored their application to Japanese texts.

Research focused on “smells” seems to have challenging aspects, but I found it to be quite interesting in many ways.

I hope this is helpful for applying these tools.