Arabic supervised learning method using N‐gram

Recently, classification of Arabic documents is a real problem for juridical centers. In this case, some of the Lebanese official journal documents are classified, and the center has to classify new documents based on these documents. This paper aims to study and explain the useful application of supervised learning method on Arabic texts using N‐gram as an indexing method (n  =  3).

The Lebanese official journal documents are categorized into several classes. Supposing that we know the class(es) of some documents (called learning texts), this can help to determine the candidate words of each class by segmenting the documents.

Results showed that N‐gram text classification using the cosine coefficient measure outperforms classification using Dice's measure and TF*ICF weight. Then it is the best between the three measures but it still insufficient. N‐gram method is good, but still insufficient for the classification of Arabic documents, and then it is necessary to look at the future of a new approach like distributional or symbolic approach in order to increase the effectiveness.

The results could be used to improve Arabic document classification (using software also). This work has evaluated a number of similarity measures for the classification of Arabic documents, using the Lebanese parliament documents and especially the Lebanese official journal documents Arabic corpus as the test bed.