mHashup

mHashup is a dynamic user interface developed by Michela Magas for use with the AudioDB database management system.




Finding duplicates in a large digital library

Administering vast quantities of music files is a major problem for large music libraries and music distributors. Tests with mHashup have shown it to be an efficient tool for administrative use with large collections.

Within a large popular music library, when choosing a segment by a famous mainstream artist like Madonna, sampling another mainstream artist like ABBA in "Hung Up", the user would be able to identify the following results:
- library duplicates of "Hung Up"
- remixes of "Hung Up"
- live versions of "Hung Up"
- the original "Gimmie Gimmie Gimmie"
- versions of "Gimmie Gimmie Gimmie"
- other songs by ABBA or Madonna which sound like the sampled shingle
- songs by other artists which sound like the sampled shingle
and most importantly:
- any DJ mix which contains this sample, and is published under a different name, without any reference to ABBA and Madonna

This allows an administrator to delete any duplicates, organise lists of versions, point to the original track, suggest similar tracks for purchase, and list all DJ recordings which have legally sampled the music. Their job is made easier and more accurate.


Detecting copyright infringement

When suspecting copyright infringement, the key is to be able to access the source of the infringement, with a precise point of origin within the plagiarised track, for exact comparison. Where recordings have been plagiarised but slightly altered by stretching or modifying the sound quality, the match may not be immediately obvious to the naked ear and one may have to trawl through hundreds of recordings looking for clues.


Helping commercial composers test melodies to prevent copyright infringement

During SIGGRAPH 2008 several composers working with the film and advertising industries approached me in view of testing their compositions against a comprehensive catalogue. Commercial composers are often at risk of breaching copyright. I was told:
" Sometimes I don't know if this catchy tune, which I am about to propose for a jingle, was something I heard in childhood or something truly original."

Testing commercial compositions before releasing them enables composers to modify their compositions if in any way in risk of copyright infringement. Composers felt mHashup was an indispensable tool in this context.


Tracing musical influences

Often a composition can be similar, but not the same as the original. All composers are influenced by others and those influences can be traced very efficiently with mHashup. A large number of users have suggested testing a named modern song against a named old composition which they had noticed to be similar. Examples included songs from Fred Astaire and Ginger Rogers movies as having a direct influence on a composition from Les Miserables. Users were eager to get their hands on a tool which would help them test their claims.

Often a tune will remind one of a melody heard long ago. With access to the relevant historic music library mHashup can locate any matching material, or even the precise track one had been trying to remember. What's more, it can pull up the metadata about a found track and in this way enable a further search through the semantic web. A musical influence can then be traced with its full history and origin.


Searching through film and video via sound

As it has been proven above, spoken word can bring results on par with a well-known melody. This enables searches through moving image media regardless of whether it contains song or word. A movie scene can be detected via a piece of dialogue or a melody. Introducing a moving-image library into the search may, therefore, bring results from different types of media.

Research has been conducted with 'Soundspotter' (Casey, M. and Grierson, M. "Soundspotter and Remix-TV: Fast Approxmate Matching for Audio-Visual Performance", in Proceedings of the International Computer Music Conference, Copenhagen, Denmark, 2007.) Further investigation is planned by integration of the existing research with mHashup.


Challenging "The Long Tail" phenomenon

When the concept of The Long Tail was published by Chris Anderson in October 2004 in Wired Magazine (http://www.wired.com/wired/archive/12.10/tail.html), it immediately hit a nerve. It addressed a new phenomenon in the music industry, overloaded with easily stored digital music files, but with no means to promote all of them. Anderson proposed that the work by lesser-known artists, who sell fewer copies of their recordings, is still economically viable in large numbers. This creates a "Long Tail" of consumers, each a fan of a virtually unknown, and 'hard-to-find' or 'non-hit' artist.

In order to benefit further from the Long Tail phenomenon, marketing relies on a crucial element of social behaviour: recommendations. Witness the following proliferation of social networking, each avidly encouraging personal recommendations. Marketing directors' work has been delegated to the general public and peer-to-peer networking.

It would be quite outside the remit of this thesis to prove the direct relationship between the publishing of the Long Tail and the subsequent success of MySpace, Facebook and Bebo. However it will be very interesting to observe the impact of a tool like mHashup on those social networking platforms. When all music becomes accessible via sound similarity rather than recommendations, the famous phrase "customers who bought this item also bought.." becomes virtually redundant.


Eliminating the generic "World Music" classification

One of the fruits of restrictive marketing tactics has been the creation of the misleading genre of "World Music". This can be seen as a giant cauldron of non-Western music styles, related only by the lack of promotion in the West. The richness of the World music heritage is hidden from all but the odd Long Tail customer while ignorance on the part of giant digital music libraries breeds further ignorance by burying it in the hard-to-find digital catalogue buckets.

mHashup can make a big difference to the status of the so called "World Music". I would argue that with any music the sound in its essence is familiar, and the context is learned. By focusing on sound only, mHashup can remove cultural barriers for its appreciation and understanding. It can further help to educate and inform about the related cultural contexts via sound.

Useful studies would be:
- Anyone coming across an interesting piece of Latin music will be at odds as to how to find similar Latin artists among thousands of names listed under "Latin". mHashup will deliver similar music and supply all the names for further searching.
- A Latin melody recorded in a track which has been catalogued as Rock or Jazz, can be matched to its origin, thus enabling a discovery of a whole new area of music
- One is able to navigate across different musical areas of the planet by connecting sounds and finding links between instruments from different parts of the world, thus spotting cultural influences and movements.


Environmental sound: matching birdsong to drills

One of the potential uses of mHashup is to match different types of environmental sound. Research has already been conducted by Oliver Bown, whilst at Goldsmiths, using the back-end tools. In a piece of public performance, Oliver matched birdsong to drills.

Environmental sound is crucial in the description of the 'atmosphere of a place'. It is possible that places with what we describe as 'good atmosphere' have something in common. mHashup can be used for this type of research, and expanded to include references to location, type of landscape/configuration, and origin of sound.


Find the riff case scenario

A music student may be able to practice a particular technique by finding examples of that same technique performed by a variety of other artists. mHashup can be used as the search facility, amplified with the addition of the corresponding musical notation and related information.


Researching sound for animation, film, and media

Often a moving-image sequence has been conceived with a particular track in mind. If this track proves to be expensive or impossible to use, another track is needed, which has the same qualities as the original.


Collecting statistics on usability

A novel way of navigating music offers opportunities for marketing analysis, study of user demographics, and statistics on the length and nature of chosen sequences.


Plug-ins for PDAs, blogs and social networking sites

A variety of user-related research could be enabled by generating plug-in versions of the tools that can be downloaded onto mobile phones, pasted in blogs and social networking sites.