By Mark Saccomano (Columbia University)
María Victoria Eyharabide is Assistant Professor of Computer Science at Sorbonne -Université and a developer on the Musiconis database project. As part of the program of seminars on the digital humanities offered to FAB-Musiconis participants, Professor Eyharabide gave an overview of database design principles necessary for efficient execution of search queries. These principles relate to indexing records in the database so that relevant results can be easily retrieved by the end user.
A simple search for a string of text (for example, using Ctrl+F to locate a word or phrase in a document file) will often return far too many records than the user requires, including many irrelevant results due to the fact that the search will look for the requested string anywhere within the record. Synonyms, homonyms, periphrasis, and the ambiguity of natural language all present further problems that limit the functionality of a such a search. Therefore, it is desirable for a database to index records in such a way that that users can find what they need quickly and easily; creating an ontology is an integral step in this process.
In an ontology, database entities are defined according to a set of criteria that allow them to be linked with other entities in meaningful ways. For example, the word harp is defined in the ontology as an element in a taxonomic classification of musical instruments. It is assigned membership in the class chordophone, which is in turn a member of the class instrument. Logical relations among classes, such as equivalency, disjunction, range and domain, are also defined through the ontology, so that, for example, “angel playing harp” is a valid search request and “harp playing angel” is not. Professor Eyharabide demonstrated how these relationships are expressed in the software program Protege.
An innovative aspect of the Musiconis database schema is the notion of a “scene” as the highest order entity modeled by the database. As an instrument is often depicted as being played by someone, often in conjunction with other players, and often with characters nearby responding to the sounds they hear, musicians and listeners are part of a single musical event, an event that gives clues to the instrument’s use and iconographic significance. The designers of the database deemed these contextual details to be a valuable and necessary component of the information they wanted to capture. So rather than simply having database entries that correspond to each image of an instrument, Musiconis records are records of scenes, of tableaux containing information about instruments, their use, their players, the surrounding figures, as well as metadata that identifies the allegorical meanings of the scenes, such as their use as visual representations of church teachings, biblical stories, or theological beliefs. Ontologically speaking, then, a record contains an abstract entity called a scene that has a set of defined attributes. These are further categorized in terms of performance data, instrument data, sound characteristics, and graphic signifiers, such as the color and formal design of the depicted scene. Such a schema provides the user with a fuller picture of the overall depth and richness of the data.