Leitmotif

What is a leitmotif?

Certain musical motives (short melodic/rhythmic/harmonic patterns) are very often used by composers as 'signposts' to the musical unfolding of the drama in a opera or other large-scale work. The leitmotifs may be identified with a certain character (sometimes being 'referred to' but not actually present on stage), an object, an emotion, or some other abstract entity that is being evoked.

The technique is particularly identified with the 19th-century composer Richard Wagner, whose operas have been continually studied for their use of leitmotifs. It is also very familiar from its use by film composers; in fact John Williams's music for Star Wars is famous for this, especially as some of the leitmotifs are essentially the same as those used by Wagner.

See Leitmotifs in Der Ring des Nibelungen and Das Rheingold, Synopsis for examples.

The use case

Many books have been published over more than a century which list Wagner's uses of leitmotifs. While he did not make public such a list himself, these lists in principle offer a kind of 'ground-truth'. However, in general the books do not attempt to include all appearances, but just the first occurrence, which is usually fairly prominent in the musical texture. So although the lists provide suitable material for queries to an MIR system, the ground truth is only partial and needs to be supplemented by the human-addition of other occurrences.

There are two sources of suitable queries:

• the books of leitmotifs (and MIDI files available via the WWW); these will essentially be symbolic queries

• the operas themselves; audio queries-by-example

Some leitmotifs, and some occurrences of all leitmotifs, will be harder to find than others. Frequently they are buried in a complex orchestral vocal texture, but more usually Wagner takes a good deal of care to ensure that the most important appearances are clearly audible. It will be especially interesting to see what is needed to maximise precision (getting occurrences of the right leitmotif high in the list) or recall (ensuring that as few as possible are missed in a manageable number of 'hits').

The Leitmotif Demonstrator

This use case idea is based on a (rejected) proposal Michael Casey and I put together for an AHRC eScience Demonstrator project last year (2006). This involved the further element of video presentation, which is not strictly necessary for this use case, but would not be so very hard to manage in principle (since in no sense would we be searching the video content, just the soundtracks).

Our brief description in the proposal was as follows:

The central idea is to index a large corpus of audio/video content, such as Wagner's operas or film scores such as the Star Wars episodes, so that visual scenes can be retrieved based on the presence of specific themes; where themes can be defined by melodies or chord progressions. Wagner?s extensive use of Leitmotif to support the dramaturgy of his operas has been widely recognized. We note that the same devices are extensively used in film to support the narrative. A theme (melody) can be selected from a visual interface presenting a list of ?theme candidates? that appear in a collection. From this list the user selects a theme, or they can vocalize a new theme by humming or singing, then, very quickly, a result screen appears with thumbnails of video scenes in which the theme appears. For the specific purpose of this demonstrator, the thumbnails are to be retrieved from multiple performances of Wagner?s operas; consisting of many tens of hours of audio and video. Clicking on a thumbnail pops up a media player that plays the scene (video) and shows its context in the overall narrative structure of the work.

There is a great deal of interest in the musicological community in the possibility of discovering new musical links between thematic materials and their use for dramatic effect in Opera or Film. Goldsmiths has jointly applied with Queen Mary for EPSRC support for a a large-scale project bringing together the UK?s leading musicologists and music information retrieval researchers to create tools and infrastructure for undertaking large- scale research into musical relationships among music collections. This project is called OMRAS2 and is an extension of the OMRAS project mentioned above. We see the role of the demonstrator as bringing together existing tools into a showcase for the future of musicology using computation. OMRAS2 will fully realize the implications of such technologies by giving access to the tools, via Grid computing and Web Services, to musicology researchers.

-- TimCrawford - 19 Jan 2007