Authors:
Amira Shoukry
and
Ahmed Rafea
Affiliation:
Department of Computer Science and Engineering, The American University in Cairo (AUC), Cairo and Egypt
Keyword(s):
Arabic Sentiment Analysis, Arabic Sentiment Lexicons, Domain-specific, Egyptian Dialect, Gulf Dialect, Arabic Opinion Mining.
Related
Ontology
Subjects/Areas/Topics:
Enterprise Information Systems
;
Recommendation Systems
;
Social Media Analytics
;
Society, e-Business and e-Government
;
Software Agents and Internet Computing
;
Web Information Systems and Technologies
Abstract:
Given the sacristy of the Arabic sentiment lexicon especially for the Egyptian and Gulf dialects, together with the fact that a word’s sentiment depends mostly on the domain in which it is used, we present SATALex which is a two-part sentiment lexicon covering the telecom domain for the Egyptian and Gulf Arabic dialects. The Egyptian sentiment lexicon contains close to 1.5 thousand Egyptian words and compound phrases, while the Gulf sentiment lexicon contains close to 3.5 thousand Gulf words and compound phrases. The development of the presented lexicons has taken place iteratively, in each iteration manual annotators analyzed tweets for the corresponding dialect to try to extract as many domain specific words as possible and measure their effect on the performance of the classification. The result are lexicons which are more focused and related to the telecom domain more than any translated or general-purpose sentiment lexicon. To demonstrate the effectiveness of these built lexicon
s and how directly they can impact the task of sentiment analysis, we compared their performance to one of the biggest publicly available sentiment lexicon (WeightedNileULex) using Semantic Orientation (SO) approach on telecom test datasets; one for each dialect. The experiments show that using SATALex lexicons improved the results over the publicly available lexicon.
(More)