Afan Oromo Text Summarization With Deep Learning Approach

Abstract

Text summarization is the technique, which automatically creates an abstract or extractive summary of a text. Text summarization is one of the research works in NLP, which concentrates on providing meaningful summary using various NLP tools and techniques. Abstractive and extractive summarizations are two methods of generating summaries from texts. This study has identified the “Afan Oromo Text Summarization in Deep Learning” as a research topic. The primary purpose of the study is to design a system and implement extractive and abstractive Afan Oromo proclamation text summarization to come up with effective and efficient summarization type as well as to evaluate the extent of the fitness of the algorithms. Hence, 583 articles of 27 Afan Oromo proclamations were used as an input data for the purpose. Accordingly, abstractive text summarization models (Sequence- 2-Sequence decoder with attention) and extractive text summarization models (TextRank) was developed for text summarization of the dataset. Different comparison measures (Rouge-1 and Rouge-2 percentage, count vectorizer, tfidf vectorizer, and soft-cosine similarity) were implemented to evaluate the text summaries produced. Results of the Rouge-1 and Rouge-2 measurement percentage index were higher for abstractive summarization than that of the extractive one in this case. Besides the algorithms and models used for both summarization methods fit for the Afan Oromo proclamation text summarizations.