MASHAEIR: Bootstrapping a Multi-Dialect Fine- Grained Emotion Thesaurus for Arabic Using Twitter
Khaled Elghamry · The Egyptian Journal of Language Engineering /The Egyptian Journal of Language Engineering · 2015
The user-generated content on social media sites, e.g. Twitter and Facebook, provides a rich source ofpeople's emotions towards products, issues, people and major events. Accordingly, the focus of more research has movedfrom negative-positive sentiment classification tasks to tasks of recognizing more fine-grained emotions. However,research on and resources for fine-grained emotion identification in Arabic texts are still lacking. To fill in this gap, thispaper introduces MASHAEIR (an Arabic word that means ‘emotions’), a corpus-based multi-dialect fine-grainedemotion thesaurus for Arabic. MASHAEIR was bootstrapped using 'big data' from Arabic Twitter from January 2007 toJuly 2015. The thesaurus is enriched with (i) different types of single- as well as multi-word terms expressing emotions,(ii) Arabic dialectal variations in the expression of emotions and (iii) scores that reflect the intensity of the emotionsconveyed through these units. The paper also presents a simple evaluation of the thesaurus coverage on a sample Twittercorpus. MASHAEIR is intended to present an outline of a large-scale and easy-to-update emotion thesaurus for Arabicthat could also be enriched in the future with more information such as gender and age preferences in expressingemotions.