In text modality, it involves detecting affect from text such as tweets, text-based conversations or an event explained using text.
ISEAR comprises of self-reported emotional situations in the form of sentences from people belonging to 37 countries on all 5 continents. It has a total of 7,665 sentences from approximately 3,000 respondents. These were labelled to one of the 7 discrete affect states.
Joy, Fear, Anger, Sadness, Disgust, Shame, and, Guilt
| Paper | Year | Metric | Code Link |
|---|---|---|---|
Emotion Stimulus comprises of sentences where a subset of the dataset has the emotional stimuli (the cause of the emotions)labelled in the text. 820 sentences in the dataset contains the cause and 1594 sentences do not contain a cause but is annotated with one of 7 seven discrete affect states.
Happiness, Sadness, Anger, Fear, Surprise, Disgust and, Shame.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
| Paper | Year | Metric | Code Link |
|---|---|---|---|
EmoInt comprises of tweets annoted one of the 4 discrete affect states and a real-valued intensity score ranging between 0 and 1. It contains a total of 7097 tweets.
Joy, Fear, Anger and, Sadness.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
| Paper | Year | Metric | Code Link |
|---|---|---|---|
SSEC comprises of tweets from SemEval 2016 Twitter stance and sentiment corpus annotated with emotion. The original dataset comprises of stance and sentiment classes. It contains a total of 4,868 tweets labelled with one, or more of, 8 discrete affect states.
Joy, Fear, Anger, Sadness, Surprise, Anticipation, Trust and, Disgst.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
DailyDialog comprises human-written multi-turn conversations with an average of 7.9 speaker turns per conversation. It contains a total of 13,118 conversations annotated with one of 6 discrete affect states.
Happiness, Fear, Anger, Sadness, Surprise and, Disgst.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
Grounded Emotions comprises of microblogs published in Twitter platform and the labels were given based on the emotion hashtags provided by the author. It contains a total of 13,118 conversations annotated with one of 2 discrete affect states.
Happy and Sad.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
Affect in Tweets comprises of tweets and this dataset was a part of SemEval-2018 Task 1. It contains a total of 10,983 tweets annotated with one or more of, the 11 discrete affect states.
Anger, Anticipation, Disgust, Fear, Joy, Love, Optimism, Pessimism, Sadness, Suprise and, Trust.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
EmpatheticDialogues comprises of multi-turn conversation comprising of a speaker and a listener. The speaker describes an emotional event in a few sentences and self-label it. The speaker then converses with a listener regarding the same. It contains a total of 24,850 conversations annotated with one of the 32 discrete affect states.
Surprised, Excited, Angry, Proud, Sad, Annoyed, Grateful, Lonely, Afraid, Terrified, Guilty, Impressed, Disgusted, Hopeful, Confident, Furious, Anxious, Anticipating, Joyful, Nostalgic, Disappointed, Prepared, Jealous, Content, Devastated, Embarrassed, Caring, Sentimental, Trusting, Ashamed and, Apprehensive.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
| Paper | Year | Metric | Code Link |
|---|---|---|---|
Happy and Sad.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
GoEmotions comprises of Reddit comments with a total of 58,009 instances. It is annotated with one or more of, the 27 discrete affect states.
Admiration, Amusement, Anger, Annoyance, Approval, Caring, Confusion, Curiosity, Desire, Disappointment, Disapproval, Disgust, Embarrassment, Excitement, Fear, Gratitude, Grief, Joy, Love, Neutral, Nervousness, Optimism, Pride, Realization, Relief, Remorse, Sadness, and, Surprise.
| Paper | Year | Metric | Code Link |
|---|---|---|---|
EmoBank comprises of english sentences which contains both the reader's and the writer's annotation of Valence (V), Arousal (A) and Dominance (D) ranging from 1 to 5. In additon to VAD annotations, a subset of this dataset have been previously annotated with 6 affect states using continous values ranging from 0 to 100. Strapparava and Mihalcea, 2007.
Continous[1,5] - Valence, Arousal and Dominance
Continous[0,100] - Anger, Disgust, Fear, Joy, Sadness and, Surprise.
| Paper | Year | Metric | Code Link |
|---|---|---|---|