Using Deep Learning to recommend Punjabi Music
| dc.contributor.author | Singh, Harpreet | |
| dc.contributor.supervisor | Tzanetakis, George | |
| dc.date.accessioned | 2018-04-26T18:33:33Z | |
| dc.date.available | 2018-04-26T18:33:33Z | |
| dc.date.copyright | 2018 | en_US |
| dc.date.issued | 2018-04-26 | |
| dc.degree.department | Department of Computer Science | |
| dc.degree.level | Master of Science M.Sc. | en_US |
| dc.description.abstract | A recommender system allows users to discover and listen to songs similar to the song they have been listening to. Collaborative filtering has been the system of choice for most music streaming services, but this type of recommendations ignore actual musical content of the songs. For genres like Punjabi music, these systems suffer from lack of historic data and hence recommendation quality is unsatisfactory. In this project, we perform experiments to determine how Convolutional Neural Networks can be used to learn musical features from Indian and specifically Punjabi Music. We investigate using these features to perform content based recommendations. We use mel-spectrograms of the songs to train a CNN on classification tasks and then use the learned weights as a feature extractor for songs. We investigate the performance of CNNs with weights trained on Western Music and fine-tuned on Punjabi Music and perform a simple study to understand the quality of recommendations. These experiments demonstrate how features learned on western music can be transferred to learn features for non-western music. | en_US |
| dc.description.scholarlevel | Graduate | en_US |
| dc.identifier.uri | http://hdl.handle.net/1828/9264 | |
| dc.language.iso | en | en_US |
| dc.rights | Available to the World Wide Web | en_US |
| dc.title | Using Deep Learning to recommend Punjabi Music | en_US |
| dc.type | project | en_US |