Evaluating the Content of LMF Standardized Dictionaries: A Practical Experiment on Arabic Language

摘要

Since the age of paper versions, dictionaries are often published with anomalies in their content resulting from lexicographer's mistakes or from the lack of efficiency of automatic enrichment systems. Many of these anomalies are expensive to manually detect and difficult to automatically control, notably with lightly structured models of dictionaries. In this article, we take advantage of the fine structure proposed by the Lexical Markup Framework (LMF) norm to investigate the detection of anomalies in the content of LMF normalized dictionaries. First, we give a theoretical study on the plausible anomalies, such as inconsistency, incoherence, redundancy, and incompleteness. Second, we detail the approach that we propose for the automatic detection of such anomalies. Finally, we report on an experiment carried out on an available normalized dictionary of the Arabic language. The experiment has shown that the proposed approach gives reasonable results in terms of precision and recall.