Muazzam Ahmed Siddiqui

Discovering the Thematic Structure of the Quran using Probabilistic Topic Model

Topic modeling refers to extracting topics from text. Topic model is a statistical model whose aim is to discover topics from a large collection of documents. A topic consists of a collection of words that are more likely to be found together in the given context of that topic or theme. This paper applies a topic model to discover the thematic structure of the Quran. For centuries, the Quran has been widely studied for the topics it contains and the relationships among them. The Holy Quran is a treasure of tremendous amount of information that addresses various aspects of human life, social as well as individual. The information present in the Quran relates in a conceptual manner although its individual bits may look unstructured and scattered. This paper attempts to use a computational method to identify this hidden thematic structure automatically. We considered each surah in the Quran as a document and used Latent Dirichlet Allocation, a probabilistic topic modeling algorithm, to discover the topics/themes. The Arabic Quran was used as the corpus instead of transliteration or translation. Our results are very promising and we were able to discover the major themes in the surahs, along with the most important terms that describe these themes.

Pre-trained Word Embeddings for Arabic Aspect-Based Sentiment Analysis of Airline Tweets

Recently, the use of word embeddings has become one of the most significant advancements in natural language processing (NLP). In this paper, we compared two word embedding models for aspect-based sentiment analysis (ABSA) of Arabic tweets. The ABSA problem was formulated as a two step process of aspect detection followed by sentiment polarity classification of the detected aspects. The compared embeddings models include fastText Arabic Wikipedia and AraVec-Web, both available as pre-trained models. Our corpus consisted of 5K airline service related tweets in Arabic, manually labeled for ABSA with imbalanced aspect categories. For classification, we used a support vector machine classifier for both, aspect detection, and sentiment polarity classification. Our results indicated that fastText Arabic Wikipedia word embeddings performed slightly better than AraVec-Web.