Exploring the music similarity space on the web

Abstract

This article comprehensively addresses the problem of similarity measurement between music artists via text-based features extracted from Web pages. To this end, we present a thorough evaluation of different term-weighting strategies, normalization methods, aggregation functions, and similarity measurement techniques. In large-scale genre classification experiments carried out on real-world artist collections, we analyze several thousand combinations of settings/parameters that influence the similarity calculation process, and investigate in which way they impact the quality of the similarity estimates. Accurate similarity measures for music are vital for many applications, such as automated playlist generation, music recommender systems, music information systems, or intelligent user interfaces to access music collections by means beyond text-based browsing. Therefore, by exhaustively analyzing the potential of text-based features derived from artist-related Web pages, this article constitutes an important contribution to context-based music information research.

A model for serendipitous music retrieval

“…To this end, we are experimenting with a set of 2.3 million tracks, for which we compute content-based similarity scores with our top-performing 3 signal-based similarity algorithm [19]. As an alternative to content-based similarity, we construct a text-based similarity measure from last.fm tags, web pages about music artists [17], and microblogs [16]. Each of these measures can be used as a similarity or diversity function.…”

An Emotion-Aware Personalized Music Recommendation System Using a Convolutional Neural Networks Approach

“…Researchers used SVM [9] and linear regression [15] to classify the songs based on the audio features of the songs [16,17]. They used traditional approaches such as the mel-frequency cepstral coefficients (MFCCs) for extracting audio features from the songs.…”

Hybrid retrieval approaches to geospatial music recommendation

“…Following the approach suggested in [13], we retrieve the top 50 web pages returned by the Bing 5 search engine for queries comprising the artist name 6 and the additional keyword "music", to disambiguate the query for artists such as "Bush", "Kiss", or "Hole".…”

“…To describe the music items at the artist level, we follow the approach proposed in [13]. In particular, we model each artist by creating a "virtual artist documents", i.e.…”

“…In accordance with findings of [10], we then use a dictionary of music-related terms (genres, styles, instruments, and moods) to index the resulting documents. From the index, we compute term weights according to the best feature combination found in the large-scale experiments of [13]: TF_C3.IDF_I.SIM_COS, i.e. computing term weight vectors and artist similarity estimates according to Equations 1, 2, and 3, respectively for tf , idf , and cosine similarity; f d,t represents the number of occurrences of term t in document d, N is the total number of documents, Dt is the set of documents containing term t, Ft is the total number of occurrences of term t in the document collection, T d is the set of distinct terms in document d, and W d is the length of document d.…”