Semantic Text Summarization for Myanmar Language using Word Embedding and Graph Based Approach

Text summarization is shortening a set of data computationally, that represents the most important or informative information within the original content. With the increment of information on the internet daily, people are expecting to read documents and texts without having to encounter unnecessary information. Text summarization of other languages have improved, but there are still many challenges for the Myanmar text summarization systems are mainly based on frequencies of words, and these systems are not equivalent in a semantic way. They cannot solve the ambiguity of the Myanmar language. The text has many meanings further the literal meaning of the words. It is very hard for computers to interpret the meaning of the words. Semantic assists in extracting important information from achieving human level understanding from the computers. This paper proposed a semantic Myanmar text summarization system that is mainly based on the integration of the word embedding model and TextRank algorithm for inter-sentence relation. Graphs are everywhere in Natural Language Processing (NLP) and a robust representation of formalism. A graph-based approach for text summarization is an unsupervised technique, is much more comfortable to visualize and is more popular than another summarization. The proposed model accuracy reached an F1 score of 0.6 on average.