Sub Event Relation Classification for Amharic Texts Using Deep Learning Approach

This research focuses on the challenge of Sub-event Relation Classification (SERC) in Amharic text, with the objective of identifying relationships between event triggers or mentions in sentences. The aim is to predict sub-event relationships in Amharic sentences, benefiting applications such as information extraction, question answering, and document summarization. Amharic text requires a lot of effort to manually identify event relationships, and there isn’t yet a model that specializes in extracting event relationships from Amharic language. Deep learning techniques, particularly LSTM and BiLSTM, are used in this work to categorize the relation type between event pairs in unstructured Amharic text in order to address this difficulty. The models are trained and assessed using a dataset of 6622 manually annotated Amharic texts. Experimental results showcase the effectiveness of the BiLSTM model, achieving a high accuracy of 93% for Amharic sub-event relation classification. The findings of this research hold significant implications for natural language processing applications in Amharic and other languages. Moreover, the developed model can be extended to other languages, making it a valuable contribution to the field of natural language processing.