Automatic Text Summarization using Maximum Marginal Relevance for Health Ethics Protocol Document in Bahasa

Abstract

The rapid development of science and technology in the health sector cannot be separated from the support of health research results. Before the research with human as a subject is conducted, every health research in Indonesia is required to make a health ethics protocol document in Bahasa which must comply with basic ethical principles. To determine whether the health research ethics protocol document has met the ethical principles, the health research ethics protocol document will be reviewed by a competent reviewer. The health research ethics protocol document consists of several parts and has a large number of pages, so to conduct a review, reviewers need a long time to understand and analyze the health research ethics protocol document. To reduce the review time, an automatic text summarization (ATS) is needed. ATS extracts important information in health research ethics protocol documents and presents it to reviewers. This research uses cosine similarity and Maximum Marginal Relevance (MMR) and TextRank to summarize the document. The MMR method is considered to have more stable results than TextRank based on the ROUGE evaluation results. The evaluation of result with the ROUGE Toolkit showed F-score value of 19.92% for document 1 and 10.98% for document 2 using MMR.