Abstract
Music has become an integral part of human life, extending its influence across various industries. For many, music is considered a necessity. With the rise of neural network technology, Music Information Retrieval (MIR) has gained prominence as a multidisciplinary field focused on processing music information and its applications. One popular approach for music captioning is the multimodal encoder-decoder architecture, which utilizes the CNN-LSTM algorithm. In this study, we develop a model that simultaneously learns from audio and text data. We explore different design choices for modality fusion, including early fusion, late fusion, and hybrid fusion, to assess their impact
![Creative Commons License](http://i.creativecommons.org/l/by-nc/4.0/88x31.png)
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Copyright (c) 2024 M. Ghazali Diarsyah, Dhanny Setiawan