TY - BOOK

T1 - The impact of online metaphors on automatic Arabic sentiment analysis

AU - Alsiyat, Israa

PY - 2025

Y1 - 2025

N2 - In this study, I investigate the unique use of Arabic metaphors in online communication. In such contexts, metaphors are often conveyed through semantic and symbolic phrases, with opinions frequently reduced to a single keyword. This concise expression suits the fast-paced nature of online interactions, where metaphorical language is commonly used to share viewpoints. As metaphor usage increases online, deeper interpretation becomes essential to uncover the sentiments behind these expressions. To explore this, I compiled the Arabic Online Metaphor Corpus (AMC), a foundational step in evaluating the impact of metaphors on sentiment. My research focuses on how Arabic online metaphors influence sentiment analysis, particularly through their semantic and symbolic nature. However, a major challenge lies in the absence of effective tools for annotating general Arabic texts. Moreover, the unique metaphorical structures found online must first be analyzed before annotation is possible. To assess AMC’s impact, we employed a state-of-the-art Arabic semantic analyzer. The limited availability of Arabic sentiment analyzers posed a significant obstacle. To address this, we used the Mazajak sentiment analyzer and additional tools tested on datasets tagged using the Arabic semantic tagger. This approach aimed to explore how metaphorical language could be analyzed and applied to sentiment classification. Our experiments demonstrated the potential of semantic annotation in accurately identifying metaphorical sentiment. We evaluated the performance of different strategies using F-score, precision, and recall metrics. As a result, this research has led to the creation of the first Arabic online metaphor corpus, an initial design for metaphor-based sentiment classification, and the evaluation of sentiment prediction tools using AMC. These contributions represent a significant advancement toward the automatic recognition and interpretation of Arabic metaphors and their associated sentiments in online discourse.

AB - In this study, I investigate the unique use of Arabic metaphors in online communication. In such contexts, metaphors are often conveyed through semantic and symbolic phrases, with opinions frequently reduced to a single keyword. This concise expression suits the fast-paced nature of online interactions, where metaphorical language is commonly used to share viewpoints. As metaphor usage increases online, deeper interpretation becomes essential to uncover the sentiments behind these expressions. To explore this, I compiled the Arabic Online Metaphor Corpus (AMC), a foundational step in evaluating the impact of metaphors on sentiment. My research focuses on how Arabic online metaphors influence sentiment analysis, particularly through their semantic and symbolic nature. However, a major challenge lies in the absence of effective tools for annotating general Arabic texts. Moreover, the unique metaphorical structures found online must first be analyzed before annotation is possible. To assess AMC’s impact, we employed a state-of-the-art Arabic semantic analyzer. The limited availability of Arabic sentiment analyzers posed a significant obstacle. To address this, we used the Mazajak sentiment analyzer and additional tools tested on datasets tagged using the Arabic semantic tagger. This approach aimed to explore how metaphorical language could be analyzed and applied to sentiment classification. Our experiments demonstrated the potential of semantic annotation in accurately identifying metaphorical sentiment. We evaluated the performance of different strategies using F-score, precision, and recall metrics. As a result, this research has led to the creation of the first Arabic online metaphor corpus, an initial design for metaphor-based sentiment classification, and the evaluation of sentiment prediction tools using AMC. These contributions represent a significant advancement toward the automatic recognition and interpretation of Arabic metaphors and their associated sentiments in online discourse.

U2 - 10.17635/lancaster/thesis/2906

DO - 10.17635/lancaster/thesis/2906

M3 - Doctoral Thesis

PB - Lancaster University

ER -