@mastersthesis{roshni2024music, author = "Vashdev, Roshni Mahtani", abstract = "The use of music to enhance the well-being of individuals can be traced back to the earliest periods of human history. At a time when mental health is a major concern and effective solutions are urgently needed, one such avenue lies in the realm of music therapy. In this project, MoodRoot has been designed, developed and evaluated. It is a web application that creates playlists that navigate between emotional states to improve mood and man- age depression and anxiety. The music recommendation system seamlessly integrates two key components: the emotional analysis of Valence (pleasantness), Arousal (activation), and Dominance (control) (VAD) features [1], and the textual analysis of song lyrics. By harnessing the power of both emotional and lyrical content, the system aims to provide personalised recommendations tailored to individual emotional states and therapeutic needs, thereby fostering positive mental health outcomes. In order to create a route of songs, the user must indicate the initial and target emotions, select the first song, and specify the number of songs in the playlist. The system’s architecture incorporates a Streamlit server for user interaction, integration with the Spotify API for playlist generation in the user’s personal account, data access components for song data retrieval, and an algorithmic calculation box. This last component uses the VAD model to map songs according to their emotion tags in order to navigate from one emotional state to another one. It also analyses the similarity between song lyrics so that the songs flow naturally. Finally, user testing and evaluation was used to assess the effectiveness of MoodRoot in inducing emotional transitions and to gather feedback for future improvements. The results of the evaluation process demonstrate the potential of MoodRoot as a tool for enhancing emotional well-being through music, and emphasise the importance of continuing to develop and refine this system to create a seamless emotional journey for listeners.", institution = "Universidad Polit{\'e}cnica de Madrid", keywords = "natural language processing;emotion analysis;machine learning", month = "June", note = "TFG", title = "{D}esign and {D}evelopment of an {E}motion-{D}riven {S}ong {R}ecommendation {S}ystem for {M}usic {T}herapy", type = "TFG", year = "2024", }