Gospel lyrics and background images retrieval system based on reflection topics

摘要

This paper presents a prototype of lyrics retrieval system that enables users to browse lyrics that relevant with the reflection topics and pictures that relevant with the lyrics. In this research, we use Christian Reflection from Santapan Rohani (http://santapanrohani.org). Santapan Rohani is the translation of Our Daily Bread (formerly RBC) devotional publications into Indonesian Language. There are many Gospel (Christian) songs that could be relevant with the topics of devotion. The lyrics that use in this research from hymn books used in Church services in Indonesia (Kidung Jemaat-KJ, Pelengkap Kidung Jemaat-PKJ, and Nyanyikanlah Kidung Baru-NKB). The text processing done in this research is tokenization, word transformation, stop words removal, and stemming. The methods used in this research are TF-IDF form term weighting, and Cosine Similarity for similarity measure. From the experiments, the precision of information retrieval must be improved. When threshold similarity=0.2 with query expansion using scripture/bible reference verses of the day, precision = 16.82%, and without query expansion, precision = 17.59% when threshold similarity = 0.25.