Afaan Oromo Multi-Label News Text Classification Using Deep Learning Approach

Abstract Classification is a technique for categorizing textual data into a form of predefined categories. Due to its major consequences in regard to critical tasks such as Information Retrieval, Question Answering, Text Summarization, and Natural Language Understanding, categorization of texts has grown in importance in many real-world applications. Text classification can be done in two different ways: manual and automatic classification. In the manual text classification, A human annotator interprets the substance of text and categorizes it suitably in classification. This procedure normally produces good results, although it is time-consuming and costly. While automatic the task of automatically assigned semantic categories to natural language text, has become one of the methods for organized online information. Despite the fact that several works are done for multi label classification in English and other languages. However, very few works have been worked for multi label classification in the Afaan Oromo language and this due the lack of resources. Most researchers have applied a machine learning approach for classification in Afaan Oromo while no researchers have used a deep learning approach for multi label classification task. Afaan Oromo news texts data are collected from different sources such OBC, FBC and BBC Afaan Oromo. We use the LSTM model using the training dataset and for the classification process. The proposed model is pass through the process of: -preprocessing, wordembedding, deep network building, training the model and classification. To predict class label, the word embedding is used separately and in various combinations through several channels of LSTM, including a single layer and a three-layer with nine outputs, Sigmoid activation functions, and binary cross entropy loss functions. we used software Python to pre-process the data and design the model. Finally, the multi label Afaan Oromo text classification model achieves an accuracy. Micro -F1 90.20% and 83% value for the best for six and nine and macro F1 90.20%and 85.41% for six and nine class respectively.