Corpus Construction and Annotation Challenge for Language Identification and Sentiment Analysis

Abstract

Corpora construction is an indispensable task in many natural language processing (NLP) domains. Thus, this paper aims to construct a linguistic resource that covers two NLP domains simultaneously, namely language identification, that distinguishes Moroccan Dialectal Arabic from Modern Standard Arabic, and sentiment analysis. It presents the construction concept, defines data sources, explains the annotation process and describes corpus characteristics. All along the paper, the description is enhanced by statistics and vivid examples. Moreover, the paper studies either language identification or sentiment analysis specificities separately.