Deep Reinforced Model dan Rules-Based untuk Peringkasan Kalimat Bahasa Indonesia
The development of technology has an impact on increasing the amount of information released the difficulty of getting information efficiently. This was strengthened by the online media Kapanlagi who claimed to make about 500 news articles per day. Therefore, this study is expected to be able to produce more and provide information in a shorter form so that it requires less time to understand information contained. This study is focused on sentences compression using Deep Reinfoced Model and Rules-Based. Deep Reinforced Model implements the Encoder Decoder algorithm and Long Short Term Memory while Rules-Based is a method for solving problems with rules that are based on knowledge. The data used in this study amounted to 1200 sentences with 3300 tokens. The results obtained from this study are sentence compression using Rules-Based method is produce a better summary seen from the value of Rouge, Rouge-1 of 49.71, Rouge-2 of 33.27, and Rouge-L of 54.33 than the summary produced by Deep Reinfoced Model with a value of Rouge-1 of 14.44, Rouge-2 of 2.82, and Rouge-L of 18.23. In addition, this study also produced a sentences compression dataset that can be used for further study.