Combining Indexing Units for Arabic Information Retrieval

Abstract

Using either stems or roots as index terms offered considerable performance to Arabic Information Retrieval (IR) systems compared to the use of surface words for indexing. Many comparative works tried to find out the best from these two indexing approaches but until then, no of the two methods widely overtook the other. Each of the two index types performed better under different test circumstances in terms of recall and precision. In this paper, the authors propose a hybrid approach combining the two indexing units in a way they take the advantages from both of them and try to overcome their shortcomings. Then, based on some combining techniques, the authors assign a weight for each indexing unit and try to find out the best weighting values.

Meta-search based approach for Arabic information retrieval

“…Even though many root-based stemmers and analyzers were proposed (Al-Kabi et al, 2015;Yaseen and Hmeidi, 2014), the most accurate tools which inspired other researchers were developed by Khoja and Garside (1999) and Ghwanmeh et al (2009). Both tools gave the best performance in many comparative works word levels (Al-Kabi et al, 2011;Al-Shawakfa et al, 2010;Ben Guirat et al, 2016;Hadni et al, 2012). They have also similar approaches for root detection.…”

“…Since the results obtained by the previously cited works demonstrated the effectiveness of combining in complex languages, some researchers tried to develop similar approaches for Arabic. To the best of our knowledge, the only research works in this field were developed by Ben Guirat et al (2016Guirat et al ( , 2019. In (Ben Guirat et al, 2016), authors proposed to combine Ghwanmeh stems, roots and verbal lemmas using ZAD collection.…”

“…To the best of our knowledge, the only research works in this field were developed by Ben Guirat et al (2016Guirat et al ( , 2019. In (Ben Guirat et al, 2016), authors proposed to combine Ghwanmeh stems, roots and verbal lemmas using ZAD collection. In Ben Guirat et al (2019), they developed new pre-indexing approaches based on 3 linear combination techniques and Kneser-Ney smoothing method.…”

“…Nonetheless, most comparative works (1) did not include Ghwanmeh stemmer which led to good performance; and, (2) did not use different corpora, especially for classical Arabic. As far as combining is concerned, previous works (Ben Guirat et al, 2016Guirat et al, , 2018Guirat et al, , 2019 used only three indexing units, thus ignoring lemmas. What is more, the most recent and popular NLP tools were not used.…”

“…Finally, we aim to combine four morphological representations (i.e. roots, stems, verbal lemmas and lemmas) rather than three (Ben Guirat et al, 2016, 2018. In fact, we plan to compare two approaches to fuse four subsystems results namely (1) linear combination as a data fusion baseline which reached the best results in previous research (Ben Guirat et al, 2016, 2018; and, (2) SVMbased meta-search which has not been previously assessed for Arabic.…”