Opinion mining from arabic quotations

We present a contribution to the Arabic Opinion Mining area with the development of a new system for quotations detection and categorization. Our approach differs from the existing work by proposing an operational tool that enables search for keywords in the categorized quotations. The categorization is performed with a rule-based system and is based on the reporting markers' meaning. The system end-user interface enables two query types within the categorized quotations: searching for keywords in the part containing eventually the quotation source (reporting clause) or in the part concerning the topic (reported clause). Our system was applied to process a corpus of Arabic newspaper articles. Both evaluations were carried out with the following results: 88 of recall and 93 of precision for the identification task; 85 or recall and a 93 of precision for the categorization task. The system is publicly accessible on the web site (http://e-quotes.net). Copyright © 2014 ACM.