Dependence-Based Segmentation Approach for Detecting Morpheme Boundaries

The unsupervised morphology processing in the emerging mutant languages has the advantage over the human/supervised processing of being more agiler. The main drawback is, however, their accuracy. This article describes an unsupervised morphemes identification approach based on an intuitive and formal definition of event dependence. The input is no more than a plain text of the targeted language. Although the original objective of this work was classical Arabic, the test was conducted on an English set as well. Tests on these two languages show a very acceptable precision and recall. A deeper refinement of the output allowed 89% precision and 78% recall on Arabic