• Home
• Research
• Publications
• Lexicons
• Labeled Data
• Invited Talks
• Contact

Arabic Sentiment Analysis and Cross-lingual Sentiment Resources

This page provides access to Arabic sentiment lexicons and sentiment annotated corpora. English sentiment and emotion lexicons are available here.

SemEval-2016 Task 7: Determining Sentiment Intensity of English and Arabic Phrases

The objective of the task is to test an automatic system’s ability to predict a sentiment intensity (aka evaluativeness and sentiment association) score for a word or a phrase. Phrases include negators, modals, intensifiers, and diminishers ­-- categories known to be challenging for sentiment analysis. Specifically, the participants will be given a list of terms (single words and multi­word phrases) and be asked to provide a score between 0 and 1 that is indicative of the term’s strength of association with positive sentiment. A score of 1 indicates maximum association with positive sentiment (or least association with negative sentiment) and a score of 0 indicates least association with positive sentiment (or maximum association with negative sentiment). If a term is more positive than another, then it should have a higher score than the other. There are three tasks corresponding to data from three domains: general English, English Twitter, and Arabic Twitter. The Arabic Twitter Sentiment Lexicon used in SemEval-2016 Task 7 includes single words and phrases commonly found in Arabic tweets. The phrases in this set are formed only by combining a negator and a word. More details can be found in the task website and the associated publication:

Semeval-2016 Task 7: Determining Sentiment Intensity of English and Arabic Phrases. Svetlana Kiritchenko, Saif M. Mohammad, and Mohammad Salameh. In Proceedings of the International Workshop on Semantic Evaluation (SemEval ’16). June 2016. San Diego, California.
Paper (pdf)    BibTeX    Presentation    Task Website SemEval-2016    Arabic Twitter Sentiment Lexicon

Arabic Sentiment (Valence) Lexicons (see this Video)

If you use any of the lexicons below, please cite the papers listed at the bottom of this page. Please see the Emotion Lexicons: Ethics and Data Statement before using a lexicon.

Automatically Created Sentiment Lexicons

These lexicons were created by measuring the extent to which the words in a tweets corpus co-occurred with a set of seed positive and seed negative terms. This is based on the idea that positive terms co-occur more with positive words and less with negative words; and negative words co-occur more with negative words and less with positive words.

Lexicon	# positive	# negative	Seeds
Arabic Emoticon Lexicon	22,962	20,342	A set of twenty three emoticons such as :) and :(.
Arabic Hashtag Lexicon	13,118	8,846	A set of 230 Arabic words that were manually selected for being highly positive or highly negative.
Arabic Hashtag Lexicon (dialectal)	11,941	8,179	A set of 483 dialectal Arabic words compiled by Refaee and Rieser (2014) from tweets.

Lexicons created by translating English Sentiment Lexicons into Arabic:

Translated Lexicon	# positive	# negative	# neutral	Original English Lexicon
Arabic translation of Bing Liu’s Lexicon	2,006	4,783	-	Link to Bing Liu’s Lexicon
Arabic translation of MPQA Subjectivity Lexicon	2,718	4,911	570	Link to MPQA Subjectivity Lexicon
Arabic translation of NRC Emotion Lexicon	2,317	3,338	8,527	Link to NRC Emotion Lexicon
Arabic translation of NRC Emoticon Lexicon	38,312	24,156	-	Link to NRC Emoticon Lexicon
Arabic translation of NRC Hashtag Sentiment Lexicon	32,048	22,081	-	Link to NRC Hashtag Sentiment Lexicon

 

Arabic Corpora Annotated for Sentiment (Valence)

If you use any of the corpora below, please cite the papers listed at the bottom of this page.

a. BBN Blog Posts Sentiment Corpus (README):

A random subset of 1200 Levantine dialectal sentences
chosen from the BBN Arabic-Dialect–English Parallel Text. The sentences are extracted social media posts and provided with their translation. Apart from manual translations, the dataset also includes automatic translations into English using the Portage MT system. We manually annotated this subset and its translations (both manual and automatic) for sentiment (positive, negative, or neutral).

This is the first such resource where text in one language and its translations into another language (both manually and automatically produced) are each manually labeled for sentiment.

b. Syria Tweets Sentiment Corpus (README):

A dataset of 2000 tweets originating from Syria (a country where Levantine dialectal Arabic is commonly spoken). These tweets were collected in May 2014 by polling the Twitter API. The dataset also includes automatic translations into English using the Portage MT system. (This dataset is not provided with manual English translation.) We manually annotated this subset and its translations (both manual and automatic) for sentiment (positive, negative, or neutral).