Gulf Arabic Linguistic Resource Building for Sentiment Analysis

Gulf Arabic Linguistic Resource Building for Sentiment Analysis Wafia Adouane1 and Richard Johansson2 {1Department of Philosophy, Linguistics and Theory of Science, 2Department of Computer Science and Engineering} University of Gothenburg wafia.gu@gmail.com, richard.johansson@gu.se Abstract This paper deals with building linguistic resources for Gulf Arabic, one of the Arabic variations, for sentiment analysis task using machine learning. To our knowledge, no previous works were done for Gulf Arabic sentiment analysis despite the fact that it is present in different online platforms. Hence, the first challenge is the absence of annotated data and sentiment lexicons. To fill this gap, we created these two main linguistic resources. Then we conducted different experiments: use Naive Bayes classifier without any lexicon; add a sentiment lexicon designed basically for MSA; use only the compiled Gulf Arabic sentiment lexicon and finally use both MSA and Gulf Arabic sentiment lexicons. The Gulf Arabic lexicon gives a good improvement of the classifier accuracy (90.54 %) over a baseline that does not use the lexicon (82.81%), while the MSA lexicon causes the accuracy to drop to (76.83%). Moreover, mixing MSA and Gulf Arabic lexicons causes the accuracy to drop to (84.94%) compared to using only Gulf Arabic lexicon. This indicates that it is useless to use MSA resources to deal with Gulf Arabic due to the considerable differences and conflicting structures between these two languages. Keywords: Modern Standard Arabic (MSA), Gulf Arabic, sentiment analysis, Arabic Natural Language Processing, Gulf Arabic sentiment lexicon 1. Introduction Arabic sentiment analysis is an active area where many works have been done recently for various topics. However, most of these works are Modern Standard Arabic (MSA) based even though most Arab people use their own dialects to express themselves online. The major obstacle to face is the non-existence of linguistic resources for Gulf Arabic, namely lexicons for sense disambiguation, sentiment lexicon, annotated data to train on, Part of Speech (PoS) tagger and above all the complexity of this dialect itself. In this paper, we try to address sentiment analysis task for Gulf Arabic using machine learning. Therefore, we started with building linguistic resources for this dialect, i.e. annotated corpus and sentiment lexicon. It is worth mentioning that there are considerable differences between these two languages and using MSA sentiment lexicon to deal with Gulf Arabic for our task has been proven to be useless, as will be shown. This paper is organized as follows: first, we give an idea of the most recent works dealing with MSA- based sentiment analysis and some attempts to deal with Egyptian Arabic for the same task. Next, we explain how we built and annotated our resources for Gulf Arabic. Then, we describe the conducted experiments and discuss the results. Finally, we describe some future directions. 2. Related Work The majority of the work done, so far, for Arabic sentiment analysis is MSA-based. This can be explained by the availability of some important linguistic resources, even though these are mostly not completely freely available for the open public. These resources include …

Content truncated to 500 words