Recomendación personalizada de canciones
Loading...
Official URL
Full text at PDC
Publication date
2024
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
Hoy en día la música es una de las representaciones artísticas más usadas y accesibles para todo el público. Es por ello por lo que cada día son más las personas que utilizan plataformas digitales que ofrecen contenido musical, como Spotify, Amazon Music o YouTube, entre otras. La amplia variedad y el aumento progresivo de canciones en estas plataformas puede hacer que sea abrumador para el usuario. Los recomendadores suponen una buena solución a este problema, siendo de gran ayuda para clasificar y mostrar al usuario contenido relacionado con sus gustos y preferencias. En este Trabajo de Fin de Grado hemos implementado un algoritmo de recomendación híbrido de canciones mediante la investigación y análisis de diferentes técnicas. En nuestro caso hemos optado por un recomendador que aproveche las ventajas de los recomendadores basados en contenido y de filtrado colaborativo atendiendo a los problemas que pueden surgir, como es el cold-start y long tail, y la confianza del usuario con la recomendación, mejorándola con la explicabilidad. Asimismo, se ha desarrollado una interfaz de usuario para mostrar visualmente los
resultados del algoritmo de recomendación al usuario.
Music is one of the most widely used and accessible forms of artistic expression today. As a result, more and more people are using digital platforms that offer music content, such as Spotify, Amazon Music, or YouTube, among others. The wide variety and the progressive increase of songs on these platforms can make it overwhelming for users. Recommenders are a good solution to this problem, being of great help to classify and show users content related to their tastes and preferences. In this Final Degree Project, we have implemented a hybrid song recommendation algorithm through the research and analysis of different techniques. In our case, we have opted for a recommender that takes advantage of the advantages of content-based recommenders and collaborative filtering, considering the problems that may arise, such as cold-start and long tail, and the user's trust in the recommendation, improving it with explainability. Likewise, a user interface has been developed to visually show the results of the recommendation algorithm to the user.
Music is one of the most widely used and accessible forms of artistic expression today. As a result, more and more people are using digital platforms that offer music content, such as Spotify, Amazon Music, or YouTube, among others. The wide variety and the progressive increase of songs on these platforms can make it overwhelming for users. Recommenders are a good solution to this problem, being of great help to classify and show users content related to their tastes and preferences. In this Final Degree Project, we have implemented a hybrid song recommendation algorithm through the research and analysis of different techniques. In our case, we have opted for a recommender that takes advantage of the advantages of content-based recommenders and collaborative filtering, considering the problems that may arise, such as cold-start and long tail, and the user's trust in the recommendation, improving it with explainability. Likewise, a user interface has been developed to visually show the results of the recommendation algorithm to the user.
Description
Trabajo de Fin de Grado en Ingeniería Informática, Facultad de Informática UCM, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2023/2024.