Arabic topic identification based on empirical studies of topic models

Abstract

This paper focuses on the topic identification for the Arabic language based on topic models. We study the Latent Dirichlet Allocation (LDA) as an unsupervised method for the Arabic topic identification. Thus, a deep study of LDA is carried out at two levels: Stemming process and the choice of LDA hyper-parameters. For the first level, we study the effect of different Arabic stemmers on LDA. For the second level, we focus on LDA hyper-parameters α and β and their impact on the topic identification. This study shows that LDA is an efficient method for Arabic topic identification especially with the right choice of hyper-parameters. Another important result is the high impact of the stemming algorithm on topic identification. Cet article met l'accent sur l'identification thématique pour la langue arabe basée sur les topic models. Nous étudions l'Allocation de Dirichlet Latente (LDA) comme une méthode non supervisée pour l'identification thématique. Ainsi, une étude approfondie de LDA a été effectuée à deux niveaux: le processus de lemmatisation et le choix des hyper-paramètres. Pour le premier niveau, nous étudions l'effet des différents lemmatiseurs sur LDA. Pour le deuxième niveau, nous nous focalisons sur les hyper-paramètres α et β de LDA et leurs impacts sur l'identification. Cette étude montre que LDA est une méthode efficace pour l'identification thématique Arabe surtout avec le bon choix des hyper-paramètres. Un autre résultat important est l'impact élevé de l'algorithme de lemmatisation sur l'identification thématique.

Analysis of online parenting community posts on expanded newborn screening for metabolic disorders using topic modeling: a quantitative content analysis

“…토픽 수 및 LDA 파라미터(parameter)가 이용되며 분석 시 연구자가 입력해야 한다. 본 연구에서는 LDA 파라미터를 선행연구[ 28 ]에 근거하여 사전 확률분포 α를 0.1, 토픽 내 사전확률분포 β를 0.01, 반복 수행 횟수를 1,000회로 설정하여 실시하였다. 토픽의 수는 통계적인 방법과 해석적인 방법을 사용하여 결정하였다.…”

Identifying Latent Topics and Trends in Premature Infant–Related Nursing Studies Using a Latent Dirichlet Allocation Method

“…12 In the LDA analysis, the researcher should set the α and β parameters and the number of iterative implementations in advance. The LDA parameters α and β have various values in many studies, but according to Naili and colleagues, 16 the α value is between 0.1 and 0.01, and the β value is 0.01. The α value represents the distribution of topics per document.…”

Topic Modeling for Amharic User Generated Texts

“…M. Naili et al [28] studied topic identification of Arabic texts using LDA. They showed that applying a stemmer increases the performance of topic identification.…”

“…The topic modeling related works are organized by approaches such as topic modeling (unsupervised) [13,14,28,[33][34][35][36], seeded topic modeling (semi-supervised) [29][30][31][32], and supervised topic modeling [17,[37][38][39].…”