Sentiment after Translation: A Case-Study on Arabic Social Media Posts

Abstract

When text is translated from one language into another, sentiment is preserved to varying degrees. In this paper, we use Arabic social media posts as stand-in for source language text, and determine loss in sentiment predictability when they are translated into English, manually and automatically. As benchmarks, we use manually and automatically determined sentiment labels of the Arabic texts. We show that sentiment analysis of English translations of Arabic texts produces competitive results, w.r.t. Arabic sentiment analysis. We discover that even though translation significantly reduces the human ability to recover sentiment, automatic sentiment systems are still able to capture sentiment information from the translations.

Enabling Deep Learning of Emotion With First-Person Seed Expressions

“…Our data collection method is most similar to Abdul-Mageed et al (2016), who also use phrase seeds to acquire tweets for Ekman's 6 basic emotions, but we extend the work to 8 emotions, expand the list of seed expressions used, improve on the manual annotation study, and empirically validate the method on the practical emotion modeling task both on our data and on an external dataset. Our work also has affinity to works on Arabic text classification (Abdul-Mageed et al, 2011;Refaee and Rieser, 2014;Abdul-Mageed et al, 2014;Nabil et al, 2015;Salameh et al, 2015;Abdul-Mageed, 2017, 2018Alshehri et al, 2018;), but we focus on emotion.…”

“…These include Mohammad (2012); Mohammad and Kiritchenko (2015); Wang et al (2012); Volkova and Bachrach (2016); Abdul-Mageed and Ungar (2017). For example, Mohammad (2012) collects a corpus of 50, 000 tweets using seed words corresponding to the 6 Ekman emotions and exploits it for building emotion models.…”

UWB at SemEval-2018 Task 1: Emotion Intensity Detection in Tweets

“…-Ultradense lexicon (Rothe et al, 2016) -LYSA Twitter lexicon (Vilares et al, 2014) 3.2.6 Arabic lexicons (L-ar): Lexicons are described in (Mohammad and Turney, 2013;Mohammad et al, 2016a;Salameh et al, 2015;Mohammad et al, 2016b).…”

Tw-StAR at SemEval-2017 Task 4: Sentiment Classification of Arabic Tweets

“…The pre-built lexica included NileULex (ElBeltagy, 2016) for MSA and Egyptian, Arabic Emotion Lexicon (Salameh et al, 2015) for emojis and Arabic Hashtag Lexicon (Salameh et al, 2015;Mohammad et al, 2016) for MSA/multiple dialects. Levantine and Gulf dialects were targeted through two manually-built lexica.…”