Multilingual Sentiment Datasets

A collection of multilingual sentiments datasets grouped into 3 classes -- positive, neutral, negative.

Most multilingual sentiment datasets are either 2-class positive or negative, 5-class ratings of products reviews (e.g. Amazon multilingual dataset) or multiple classes of emotions. However, to an average person, sometimes positive, negative and neutral classes suffice and are more straightforward to perceive and annotate. Also, a positive/negative classification is too naive, most of the text in the world is actually neutral in sentiment. Furthermore, most multilingual sentiment datasets don't include Asian languages (e.g. Malay, Indonesian) and are dominated by Western languages (e.g. English, German).

For emotions related datasets, I group the negative (respectively positive) emotions into the negative (respectively positive) class. For ratings datasets I assign the 1 star reviews to the negative class, 3 star review to the neutral class and assign the 5 star review to the positive class.

Disclaimer: All credits goes to the respective dataset owners, this repository is merely an aggregation of the datasets.

Datasets

Dataset name	Language	Source	No. of texts	Classes
IndoNLU (EmoT)	Indonesian	Twitter	train: 3521 val: 440 test: 440	anger, fear, happy, love, sadness
IndoNLU (SmSA)	Indonesian	Online platforms	train: 11000 val: 1260 test: 500	positive, negative, neutral
IndoNLU (CASA)	Indonesian	Automobile platforms	train: 810 val: 90 test: 180	positive, negative, neutral (6 aspects)
IndoNLU (HoASA)	Indonesian	Hotel reviews	train: 2283 val: 285 test: 286	positive, negative, neutral, positive-negative (10 aspects)
Multilingual Amazon Reviews	English, Japanese, German, French, Chinese, Spanish	Amazon	For each language: train: 200,000 val: 5,000 test: 5,000	1 star, 2 star, 3 star, 4 star, 5 star
GoEmotions	English	Reddit	211225	admiration, amusement, anger, annoyance, approval, caring, confusion, curiosity, desire, disappointment, disapproval, disgust, embarrassment, excitement, fear, gratitude, grief, joy, love, nervousness, optimism, pride, realization, relief, remorse, sadness, surprise
Offenseval Dravidian	Tamil-English, Malayalam-English, Kannada-English	Social media	Tamil: train: 35139 val: 4388 Malayalam: train: 16010 val: 1999 Kannada: train: 6217 val: 777	Not_offensive, Offensive_Untargetede, Offensive_Targeted_Insult_Individual, Offensive_Targeted_Insult_Group, Offensive_Targeted_Insult_Other, not-{lang}
SemEval-2018 Task 1: Affect in Tweets	English, Arabic, Spanish	Twitter	English: train: 6838 val: 886 test: 3259 Spanish: train: 3561 val: 679 test: 2854 Arabic: train: 2278 val: 585 test: 1518	anger, anticipation, disgust, fear, joy, love, optimism, pessimism, sadness, surprise, trust
Emotion	English	Twitter	train: 16000 val: 2000 test: 2000	anger, anticipation, disgust, fear, joy, sadness, surprise, and trust
IMDB	English	Movies	train: 25000 test: 25000	positive, negative
Amazon Polarity	English	Amazon	train: 3600000 test: 400000	positive, negative
Yelp Reviews	English	Yelp	train: 650000 test: 50000	1 star, 2 star, 3 star, 4 star, 5 star
Yelp Polarity	English	Yelp	train: 560,000 test: 38,000	positive, negative