COMPARATIVE STUDY OF TOPIC SEGMENTATION ALGORITHMS BASED ON LEXICAL COHESION: EXPERIMENTAL RESULTS ON ARABIC LANGUAGE

Topic segmentation is essential for a lot of Natural Language Processing (NLP) applications, such as text summarization or information extraction. The objective of this research is to evaluate the effectiveness of topic segmentation algorithms in identifying the thematic breaks in Arabic texts. For this aim, a group of 7 readers are asked to identify the changes of theme that they discerned in 5 Arabic texts of different domains. The resulting judgments are used to evaluate the relative performance of two of the main algorithms of segmentation proposed in the literature: C99 and Text Tiling, using the classical Recall/Precision evaluation metrics and the recently introduced Reader Judgment method. The experimental results show that with only a few improvements, existing algorithms for segmenting English texts are also efficient for segmenting Arabic texts.