SENTENCES SIMILARITY DETECTION IN INDONESIAN POETRY COMPARISON USING SIAMESE MALSTM

Poetry is one of the fashions for someone to express their thoughts and feelings beautifully and imaginatively. To preserve a better transformation in the evolution of poetry, there must be a critical thinking for something new and original. Comparative literature is one of the methods often used to examine the originality of poetry. In comparative literature, the similarities and differences in several poetry become the object of research. One that can be compared is the semantic meaning present in each line of poetry. In this study, the originality verification will be carried out by the model of machine learning which is trained to understand the semantic meaning and similar wording of each line of poetry pairs. This study uses the Siamese MaLSTM algorithm to create the model needed. The identifier model is built using Indonesian poetry collection and goes through several stages such as data preprocessing, word embeddings, training and testing. The result of the evaluation model provides a good accuracy in recognizing semantic similarity in sentences pairs. This study also presents the description of system application which needed to gather Indonesian poetry collection, identify the similarity of poetry from the prior and manage Indonesian poetry community. The machine learning model that has been built will be used in the system application as a cloud service that will check the originality of registered poems with the poetry collection in the database.