This site uses AI for smell information extraction and image generation. Please note that it may contain errors or inaccuracies. Learn more

Emotion Inference

This database automatically infers the emotions evoked by smell descriptions found in The Tale of Genji. This page explains how this inference system works.

Theoretical Foundation: Plutchik's Wheel of Emotions

Our emotion classification is based on psychologist Robert Plutchik's "Wheel of Emotions." This model defines eight basic emotions, each with three intensity levels.

The Eight Basic Emotions

EmotionWeakModerateStrong
JoySerenityJoyEcstasy
TrustAcceptanceTrustAdmiration
FearApprehensionFearTerror
SurpriseDistractionSurpriseAmazement
SadnessPensivenessSadnessGrief
DisgustBoredomDisgustLoathing
AngerAnnoyanceAngerRage
AnticipationInterestAnticipationVigilance

Compound Emotions (Dyads)

Combinations of basic emotions create compound emotions:

  • Love: Joy + Trust
  • Awe: Surprise + Fear
  • Remorse: Disgust + Sadness
  • Contempt: Anger + Disgust

How the Inference Works

Keyword Matching Method

To infer emotions from smell descriptions, we use a keyword matching approach. This method detects whether specific emotion-related keywords appear in the text.

Analyzed Fields

Three fields from each smell entry are analyzed:

  1. Effect: The emotional impact or reaction the smell causes
  2. Quality: The qualitative characteristics of the smell (good/bad, strong/weak, etc.)
  3. Circumstances: The context or scene where the smell is perceived

Inference Examples

Example 1: An Unpleasant Smell

"From the hems of the attendants' robes came an unbearable, unpleasant odor."

The expression "unpleasant" (まさなき in the original) triggers the inference of disgust and annoyance.

Example 2: A Pleasant Smell

"The fragrance of plum blossoms mingled with the scent from within the blinds, like the pure land of a living Buddha."

Expressions of beauty and pleasantness trigger the inference of joy and serenity.

Classical Japanese Emotion Expressions

The Tale of Genji contains unique emotional expressions that differ from modern Japanese. These classical terms are included in our inference system.

Classical ExpressionModern MeaningInferred Emotions
あはれ (aware)Deep emotional resonanceEmpathy, Sadness
もの哀し (mono-awashi)Vague melancholySadness, Pensiveness
つれづれ (tsurezure)Idle boredomBoredom, Pensiveness
なつかし (natsukashi)Feeling of fondnessLove, Trust
かしこ (kashiko)Awe-inspiring reverenceAwe, Fear

Emotion Filtering Feature

On the smell list page, you can filter smells by their inferred emotions:

  1. Select emotions from the facets on the left side
  2. Only smells with the selected emotions will be displayed
  3. Selecting multiple emotions shows smells that have all selected emotions (AND condition)

Technical Details

For more detailed technical information, please refer to:

RDF Data

The emotion vocabulary is also available in RDF (Turtle format). It is compatible with the Odeuropa project and follows the SKOS (Simple Knowledge Organization System) standard.

Future Developments

Currently, we use keyword matching for emotion inference. Future enhancements under consideration include:

  • Advanced LLM-based Emotion Analysis: Using large language models for deeper contextual understanding
  • Emotion Network Visualization: Creating network graphs connecting smells that share common emotions
  • Narrative Emotion Progression: Visualizing how emotion distributions change across different volumes of the tale