Peringkasan Dokumen Berbahasa Indonesia Berbasis Kata Benda

Abstract

Manual text summarization requires time and expertise, therefore automatic summarizing is preferred. This research aims to compare the results of noun-based summary on four experiments (1) PageRank based on content overlap similarity measure, (2) PageRank based on cosine similarity measure, (3) title based on content overlap similarity measure, and (4) title based on cosine similarity measure. In the first and second experiment, the ratings were done by the PageRank algorithm. Evaluation of summaries use sis judges, ie. three Indonesian high school teachers and three general readers. Their judgements were compared using the kappa measure. The results of each experiment given by three judges from Indonesian high school teachers obtained an average relevance rating 100%. For general readers, the value of relevance are various percentage levels with the highest value in third experiment is 82.79% while the lowest is in second experiment with a value of 65.59%. Summarization on title based experiment is better than PageRank based.