Claudio Baccigalupo | claudio @ iiia . csic . es

Institut d'Investigació en Intel·ligència Artificial
Campus de la UAB, 08193 Bellaterra, Spain.
Tel: +34.935809570 (259), Fax: +34.935809661

My life

Claudio Baccigalupo

I was born 30 summers ago in Milan, Italy. During these years I have been playing bass guitar, deploying Web applications and a PS2 videogame, learning English, French, Spanish and a bit of Catalan, changing residence nine times, getting married, changing the colour of my hair too many times, completing a 5-years degree in Computer Science, programming on every platform and language, investigating human-machine interaction, driving a yellow Fiat Seicento, playing in several bands, sleeping, eating, drawing, travelling, drinking, writing, loving and most of all listening to music.

Since June 2005, I am a PhD student in the IIIA, under the supervision of Enric Plaza.

My research project

My research focuses on recommending systems in a musical context.

[2009 : PhD is over]

After four intense years of work, I have finally delivered my thesis. The discussion has taken place at the IIIA on November 6th, 2009.

Poolcasting is an intelligent technique to customise musical sequences for groups of listeners. Poolcasting acts like a disc jockey, determining and delivering songs that satisfy its audience. Poolcasting additionally ensures that the played sequence does not repeat the same songs or artists closely and that pairs of consecutive songs ‘flow’ well one after the other, in a musical sense.

The thesis describes the poolcasting technique as well as Poolcasting Web radio, an innovative online radio system that broadcasts music programmes customised in real time for their listeners. A set of experiments are reported that evaluate how much the size of the group and its musical homogeneity affect the performance of poolcasting.

[2008 : Social Genre Affinity]

In most organisational schemas (record stores, Web pages), musical artists have a unique genre label attached (Pop, R&B, Electronic, …). However, most artists have affinity to multiple genres—for instance, Madonna. Indeed, most listeners see genres as fuzzy categories (Madonna is very Pop, a bit R&B, partly Electronic), more than as exclusive, Boolean categories (Madonna is Pop, is not R&B nor Electronic).

Since 2008, part of my research is to provide methods to uncover these fuzzy artists-genres affinities, based on how differently people make use of these artists and genres. For instance, in the case of Madonna, some of her songs are played in Pop bars, others can be heard in R&B clubs, others are mixed with Electronic themes, and so on.
The first results of this research, focused on the co-occurrence analysis of artists and genres in playlists, are described in the paper presented in September 2008 at the International Symposium on Music Information Retrieval (ISMIR '08):

More details on the analysis and results are also available in the Affinity Labs page.
The importance of the proposed analysis is that it can be applied to any domain that presents objects (e.g., artists), categories (e.g., genres), and social behaviour data (e.g., multiple human-compiled playlists), in order to uncover a more detailed and “social-aware” knowledge on that domain. Our world is full of categorisations (musical tags, movie genres, market types, human conditions, car ranks, fashion styles); my position is that the best way to understand how people ‘see’ and ‘feel’ these categories is to observe and analyse how people ‘use’ objects from these categories, under the assumption that when two objects or categories co-occur closely and often, they have some form of shared cultural affinity that is worth investigate.

[2007 : Social Web Radio]

A social Web radio is a virtual environment where participants listen to recommended songs, share their music, comment with similar users, and get a group-customised musical selection. Since 2007, I am working on a novel Web radio architecture called Poolcasting which implements this idea combining several different AI techniques (CBR, Pattern Mining).

A broad description of the CBR technique used to schedule the songs to play on each channel is included in the paper presented in August 2007 at the International Conference on Case Based Reasoning (ICCBR '07), where it was granted the Best Application Paper Award:

A prototype version of the radio, currently running in our Institute, was first presented in August 2007 at the International Computer Music Conference (ICMC '07) in Copenhagen, and described in the paper included in the ICMC proceedings:

The musical knowledge that drives the automatic selection of songs played in Poolcasting comes from a set of playlists collected from the users of MyStrands. The data mining technique employed to infer song and artist associations from the cooccurrences of tracks in playlists is described in the paper presented in September 2007 at the Web Mining 2.0 Workshop, part of the ECML/PKDD '07 Conference that took place in Warsaw:

As of December 2007, I am working on the evaluation of a Poolcasting Web radio currently tested in our Research Institute by a few users.

[2006 : Automatic Playlist Generator]

Before focusing on Web radios, I investigated how to apply Case Based Reasoning for recommending playlists.

I focused on the retrieval of contiguous sequential patterns of songs to build a comprehensive domain knowledge about songs that can sound well together in a playlist. This approach can be employed in any context where information is available in a sequential form and where the order is important for the elements in the sequence (as in playlists), and is described in the paper presented in September 2006 at the European Conference on Case Based Reasoning (ECCBR '06):

In May 2006, I deployed an online playlist recommender on MyStrands Labs Web-page, exploiting the enormous database of playlists (about 300,000) published by the users of the MyStrands music community. A debug version is available as well, less user-friendly, but with more parameters that can be tuned to best personalise the recommendation.

[Keynote Presentations]

Should you wish to know more about the work I have been doing so far, and the future directions that I will explore, take a look at the presentations I made during the last years:

My past experience

In December 2003, I graduated from the Università degli Studi di Milano with a thesis on the integration of audio technologies in video games, partly describing the the work I carried out as an Xbox audio programmer for the game Ruff Trigger. The first part of the thesis was later translated to English:

I am quite new to scientific research, and I am eager to know everyone's opinion about my work. Please send all of your feedback to my e-mail; I will happily read your comments, and reply to everyone. Thanks, Claudio :o)