Extractive Myanmar News Summarization Using Centroid Based Word Embedding

Abstract

Nowadays, many researches are going on for text summarization because there are a lot of data on the internet and it is required to process, store and manage. Text summarization is a process of distilling important information from the original text and presents that information in the form of summary. The system is proposed to summarize Myanmar news with centroid based method. Centroid based method ranks the sentences based on their similarity to the centroid. Centroid based method uses the bags of words model to represent sentences. Bags of words representation does not capture the semantic relationship between words. To overcome this problem, centroid based method is combined with word embedding representation instead of bags of words in this paper. Experiments were done on Myanmar news dataset. Centroid based on word embedding method gets better performance than centroid based on bags of words method.