Themefinder provides search access to a database containing tens of thousands of short melodies and musical incipits. Themefinder has been developed at the Center for Computer Assisted Research in the Humanities at Stanford University and the Cognitive and Systematic Musicology Laboratory at the Ohio State University.
Themefinder can be accessed at: http://www.themefinder.org
The Themefinder database contains the initial portions of melodies for about 20,000 themes from Classical and Folksong repertoires. When searching for recordings and scores of particular works, we usually rely on title information. When title information is not available but melodies can be recalled, they can be searched using Themefinder.
Text searches normally contain two kinds of variable elements: known letters and unknown letters. A text search such as run* would cover all possible combinations of plurals, verb forms and the like of the word run[run = a combination of known letters and * = any combination of unknown letters].
Musical searches using Themefinder may be conducted at a series of levels of specificity. Those with perfect pitch or an exact recollection of a melody will be able to construct queries that search precisely for the notes specified using the pitch procedure. Those with only a general idea of a melody may search at several more general levels. These melodic generalities include:
Other procedures are available for refined queries.
All data in Themefinder databases are encoded in the Humdrum **kern format. David Huron, the originator of the Humdrum Toolkit, which is a set of Unix programs for musical analysis, began encoding musical incipits and developing the search engine for Themefinder in 1996. These underlying data files are viewable by clicking on the "H" buttons. such as this example: .
The notated incipits which are displayed with the search results are generated by software created by Walter B. Hewlett. They are processed through the MuseData system and are presented on the Web as GIF images. Displayed themes/incipits are truncated at the end of one musical system. These examples can by pasted onto a clipboard developed by Craig Sapp, as shown in this example: . The clipboard page can save sets of themes found while searching Themefinder and can be saved as an HTML document to be viewed later.
All of the MIDI files for available incipits have been converted from ASCII data in the **kern format.
Andreas Kornstädt designed the initial interface for Themefinder. This design has been extended and adapted to current needs by Craig Stuart Sapp with the assistance of Bret Aarden. CGI scripts written in the PERL programming language collect the forms generated by the search page. The search pages convert the user's query into a format which can be run with the Humdrum thema command to search the thematic databases for matches. Resulting matches are displayed as graphical notation with brief reference information (title, composer or repertory). Buttons next to the query form give detailed information on the appropriate construction of queries. Buttons below the incipits also give access to the primary **kern datafile contents as well as MIDI playback.
In the first phase of the work, incipits were encoded in MIDI format, but this proved inadequate for enharmonically sensitive searches. Walter B. Hewlett enhanced the original classical dataset to facilitate accurate conversion to graphical display. More recently, data have been encoded directly in the **kern format or converted from other ASCII formats to **kern, such as the Essen folksong database originally encoded in the EsAC format. More recent phases of the data entry work have been carried out by music students including Leigh VanHandel, Zoë Chafe and Maria Heifetz.