Hybrid Approach to Word Sense Disambiguation for Hadiyyisa Language Using Supervised Machine Learning Models

Abstract

Word sense disambiguation is a core problem in many tasks related to natural language processing. The absence of automatic word sense disambiguation for Hadiyyisa language can be a challenge for the development of natural language processing applications such as text retrieval system, text processing, and automatic sentence parser. In addition to that, some challenges hamper the effectiveness of WSD applications for the Hadiyyisa language, such as ambiguity, the stop word identification problem, and the limitation of digital datasets. As a solution to this, a word sense disambiguation model has been developed for Hadiyyisa language to address the problem of lexical ambiguity. To conduct the research, a lexical sample tagged datasets for two senses of an ambiguous word are considered and used to evolve an optimized hybrid model that correctly disambiguates the sense of the given word considering the context in which it occurs. A new model has been developed by using supervised machine learning models and proposed model have been compared with several existing models. The performance is evaluated with their accuracy, precision, recall, and f1- score including the confusion matrix considered for accuracy comparison. The performance evaluation results revealed that the proposed Hybrid (SVM-NB-NN) model gives better performance for classifying the correct senses of classes than existing models.