Enhancing Performance of Hybrid Named Entity Recognition for Amazighe Language

Named Entity Recognition (NER) involves the identification and classification of named entities in texts. This is an important subtask in most high level NLP applications and semantic Web technologies. Besides, various studies have been done on NER for most of the languages and in particular for English. However the studies for Amazighe have lagged behind these for a long while. Recently, Amazighe NER have caught more attention due to the increasing flow of Amazighe texts available on the Web and the need to discover named entities occurring in these texts, considering the fact that a difference in language impose new challenges. Some systems using different approaches have been proposed in terms of extracting Amazighe named entities, however the recent system proposed based on a hybrid approach, the only existing hybrid system, reports a drop in F-Measure from 93 to 73% when compared to the rule based approach. In this paper, we present our enhancement of the previously proposed method by adding a new set of handcrafted lexical resources and a new set of features. The system is able to identify seven different kinds of entities such as “Person”, “Location”, “Organization”, “Numbers”, “Percent”, “Money”, “Date/Time”, it was tested on our Amazighe corpus “AMCorp” with satisfactory results.