Rule Based Part of Speech Tagger for Arabic Question Answering System

Abstract

AbstractPart of Speech (POS) Arabic wording is difficult to read in detail and its functionality affects many programs and activities in the Natural Language Processing (NLP) area. POS tagging is a process to assign POS such as a verb, adjective, adverb, noun in each word for any sentence. Farasa is an active and reliable text processing toolkit for Arabic documents. It is an assortment of Java libraries and CLIs for MSA.2. These incorporate a separate tool for Arabic text Diacritizer, segmentation/tokenization module, POS tagger, Named Entity Recognition (NER), and parsing. One of the limitations over the Farasa affects the post-processing results due to the presence of inappropriate tags. For our application on question answering system(QAS) correct POS, tagging is essential for better accuracy. The POS tagger is developed using a rule-based approach which is based on domain-specific. The corpus (database) on which the rule-based POS tagger is built is centered on the core subject of the Arabic language 4th standard textbook of Arabic Medium state board of Yemen. During the development of QAS, the POS tagger is a very essential stage in which the answers for the framed questions are obtained from the paragraphs of a given lesson. The present article provides insights into the complete process of linguistic rule-based POS tagger development for QAS. Sentence segmentation, word tokenization, to stemmer development which becomes an important part of proper morphological analysis is explained. As a result, the morphological analyzer is the input to the rule-based POS tagger. Ultimately, in this article, a comparison of marking based on our POS rule with Farasa is presented and for QAS, our rule-based POS tagger gave better results than Farasa.KeywordsArabic languageNatural language processing (NLP)Part of speech (POS) taggingMorphological analyzerStemmerTokenization