Detect Mentioned Emotion

Detect emotions that user mentioned in conversation

This documentation will guide you in identifying emotions verbalized in conversations. It provides an overview of how to use the emotion-related classes. Note that we do not perform emotion detection, but rather identify emotions that users mention in their conversations. For instance, if a user says "I feel happy," we detect the word "happy" as an emotion.

Detection of mentioned emotions is available in 🇺🇸 English, 🇨🇿 Czech, 🇩🇪 German, and 🇪🇸 Spanish dialogues

Overview

Flowstorm allows you to detect various emotions within text using a comprehensive list of predefined emotions. These emotions are hierarchically structured into emotion classes and subclasses. You can access the detected emotion using a dedicated function provided in the Input class.

Examples of Usage

The Input class provides the following two functions to access the detected emotions:

Use the input.entity() function to access the first detected emotion of a specific type:

val emotion = input.entity("Emotion")

Use the input.entities() function to access all the detected emotions:

val emotions = input.entities("Emotion")

You can also access attributes of emotion:

val emotion = (input.entity("Emotion").value as ai.flowstorm.core.model.Emotion)

You can then use the detected emotion to fine-tune the conversation flow or generate more context-aware responses. For example, by taking the detected emotion into account, an AI conversational agent might respond positively to joyful emotions:

val emotion = (input.entity("Emotion").value as ai.flowstorm.core.model.Emotion)
if (emotion.subclass == EmotionSubclass.JOY) {
    toRespondPositively
}

Or, the agent could provide comfort if detecting sadness:

val emotion = (input.entity("Emotion").value as ai.flowstorm.core.model.Emotion)
if (emotion.subclass == EmotionSubclass.SADNESS) {
    toProvideComfortingWords
}

To detect an emotion and access its properties, follow this example:

First, ensure that the conversation context is set up in Flowstorm. Then, use the Input class to detect the emotion from user input.

val detectedEmotion = (input.entity("Emotion").value as ai.flowstorm.core.model.Emotion)

After detecting the emotion, access its emotion class, noun, adjective, arousal, and valence using the emotion object's properties:

// Access emotion class
val emotionClass = detectedEmotion.emotionSubclass.emotionClass

// Access noun
val emotionNoun = detectedEmotion.noun

// Access adjective
val emotionAdjective = detectedEmotion.adjective

// Access arousal
val emotionArousal = detectedEmotion.arousal

// Access valence
val emotionValence = detectedEmotion.emotionSubclass.valence

Now, you have access to the detected emotion's properties, which can be used to guide the conversation flow, generate context-aware responses, or for any other purpose that requires emotion understanding:

logger.info("Emotion Class: $emotionClass")
logger.info("Emotion Noun: $emotionNoun")
logger.info("Emotion Adjective: $emotionAdjective")
logger.info("Emotion Arousal: $emotionArousal")
logger.info("Emotion Valence: $emotionValence")

Emotions in Flowstorm

Flowstorm offers a broad range of emotions in the form of an enum type called Emotion. These emotion types consist of various attributes such as:

• Noun

• Adjective

• Emotion subclass

• Arousal type

• Valence numeric value

• Arousal numeric value

Emotion Types and Descriptions

Each emotion enums holds a noun, an adjective, and an emotion subclass base, such as:

Emotion	Noun	Adjective	Subclass	Arousal	Valence Numeric	Arousal Numeric
ALL_RIGHT	all right	all right	JOY	NORMAL	-	-
AMAZINGNESS	amazingness	amazing	JOY	HIGH	7.72	6.05
AMUSEMENT	amusement	amused	AMUSEMENT	HIGH	7.00	4.82
AT_EASE	easiness	at ease	RELIEF	NORMAL	-	-
AWE	awe	awed	ADMIRATION	HIGH	6.85	3.83
AWESOMENESS	awesomeness	awesome	EXCITEMENT	HIGH	7.86	6.05
BETTER_MOOD	better mood	better	RELIEF	NORMAL	-	-
BLESSEDNESS	blessedness	blessed	GRATITUDE	NORMAL	7.50	3.55
BLISS	bliss	blissed	JOY	HIGH	7.62	4.13
BLISSFULNESS	blissfulness	blissful	JOY	HIGH	7.67	3.74

Emotion Subclasses

Each emotion comes with an assigned emotion subclass that represents the emotion's broader context:

Emotion Subclass	Emotion Class	Valence
ADMIRATION	JOY	POSITIVE
APPROVAL	JOY	POSITIVE
GRATITUDE	JOY	POSITIVE
AMUSEMENT	JOY	POSITIVE
PRIDE	JOY	POSITIVE
OPTIMISM	JOY	POSITIVE
JOY	JOY	POSITIVE
EXCITEMENT	JOY	POSITIVE
RELIEF	JOY	POSITIVE
LOVE	LOVE	POSITIVE

Emotion Classes

The Emotion Class serves as the higher-level categorization for specific emotions within the Flowstorm framework and helps to further group emotions into broader contexts. In the following table, you'll find various Emotion Classes with their respective values:

Emotion Class

UNSPECIFIED

LOVE

JOY

SADNESS

ANGER

FEAR

DISGUST

SURPRISE

SHAME

GUILT

Arousal and Valence

Arousal is an enum that describes how intense an emotion is, and Valence is an enum that describes whether an emotion is positive, negative, or ambiguous:

Arousal	Valence
UNSPECIFIED	UNSPECIFIED
LOW	NEGATIVE
NORMAL	AMBIGUOUS
HIGH	POSITIVE