Markus Schedl, from the Department of Computational Perception of Johannes Kepler University in Linz (Austria) will give a series of lectures on context-based Music Information Retrieval.

* 23.02: 15:30-16:30 (room 52.321) DTIC Research seminar on "Geo-Aware Music Information Extraction from Social Media"

Abstract: The abundance of data present in Social Media opens an unprecedented source for information about every topic of our daily lives. Since music plays a vital role in many persons' lives, information about music items is found in large amounts in data sources such as social networks or microblogs. In this talk, I will report on latest findings in Social Media Mining to extract meaningful musical information from microblogs. Specifically, I will address the topics of similarity measurement, popularity estimation, and cultural-aware music taste and trend detection. In addition to elaborating on the methodological background, I will present some application scenarios and demonstrator systems that strive to illustrate some application domains of this interesting research field.

* 28.02-01.03 (room 52.S27) Lectures on: "Context-based Music Information Retrieval"

These lectures give an introduction to Music Information Retrieval (MIR), with a focus on context-based methods. MIR is concerned with the extraction, processing, and use of various music-related information from various musical data sources (scores, digital audio, live concerts, collaborative tags, video clips, album covers, etc.). I will focus on feature extraction (context- and Web-based), similarity measurement, and applications of MIR. I will also strive to include my latest research on Social Media Mining for MIR.

• 28.02.: 12:00-14:00 Introduction to MIR

What is MIR? - definitions, important key aspects, subfields and typical tasks, basic scheme of an MIR system, basics in different retrieval approaches, feature extraction (audio and contextual), and similarity measurement

• 29.02.: 12:00-14:00 Context-based Feature Extraction

motivation, data sources for contextual features, specific biases and problems of contextual features, term vector-based (web terms, tags, lyrics) and co-occurrence-based (play lists, page counts, P2P networks) approaches

• 01.03.: 10:00-12:00 Similarity Measurement and Applications

similarity measurement on different kinds of music-related data (from scalar to multi-instance, multi-dimensional data), selected applications developed by the Department of Computational Perception / Johannes Kepler University, Linz, Austria (for instance, user interfaces to music)

Markus Schedl's website