Beta 1


Title Cognitive semantics of lyrics and audio features in songs
Author Koclega, Kamil Krzysztof
Supervisor Larsen, Jakob Eg (Cognitive Systems, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Petersen, Michael Kai (Cognitive Systems, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Mørup, Morten (Cognitive Systems, Department of Informatics and Mathematical Modeling, Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark)
Institution Technical University of Denmark, DTU, DK-2800 Kgs. Lyngby, Denmark
Thesis level Master's thesis
Year 2010
Abstract Temporal and time-invariant analysis of songs' lyrics and audio features are common research areas of Music Information Retrieval Scientists. There are numerous research groups deeply analyzing different sound features or using variety of Natural Language Processing algorithms to analyze lyrics. This thesis follows other researchers' steps by looking at songs' lyrics and sound features. However, at the same time it moves even further and try to find a correlation between both data sets in time-invariant analysis. The group of songs get selected based on a common criteria - emotional load, which is recognized using Latent Semantic Analysis. Afterwards, Chroma features get extracted from selected songs' music. After having both data sets ready, Nonnegative Matrix Factorization and Principal Components Analysis is run against a single data set to finish with Canonical Correlation Analysis being run on both data sets simultaneously. Each of techniques being run on data reveals hidden patterns among a single data set as well as between both of them.
Imprint Technical University of Denmark (DTU) : Kgs. Lyngby, Denmark
Series IMM-M.Sc.-2010-91
Fulltext
Original PDF ep10_91_net.pdf (0.54 MB)
Admin Creation date: 2010-11-02    Update date: 2011-06-10    Source: dtu    ID: 268519    Original MXD