search

Blog

August 9, 2013

Toward Less Precious Cataloging

For a number of years after college I was in the workers compensation and general liability industry, bouncing around among positions from file clerk to claims adjuster to data analyst. Somewhere in there I did transcriptions of phone interviews between adjusters and claimants. One of the more depressing parts of the job, I tried to distract myself by attempting to transcribe things as faithfully as possible, reflecting every single er, uhm, stutter, pause, and garbled sentence fragment. I wanted to reflect how people really spoke in all of its confusion and banal glory.

Also, coming out of literary studies, I forced myself to imagine there was some meaning to those various tics and patterns, but I also wanted to see how far I could stretch the plasticity of punctuation. What length of pause was a comma, a dash, a double (or triple!) dash, an elipsis? Are fragmented thoughts governed by semi-colons even if such marking was unintended by the speaker? One almost has to feel (or, at least I do) that under copyright law my expression of this recorded speech was so transformative that those transcriptions were wholly new works of art!

Did I mention I was very bored at this job?

It is true, however, that transcription is an art. It is the interpretation of information in one medium into another, not necessarily compatible, medium. I guess it wouldn’t be considered a fine art because it is more along the lines of didacticism — the direct communication of that information with maximal understanding as a goal, rather than an approach of impressionism, nuance, and open questions of interpretation.

But an art it is, requiring specialized knowledge to be done well and skill greater than just typing the words as they come. One can look at the frequent mistakes of spell check and speech-to-text, and consider the complex set of rules structured around such utilities, to appreciate how difficult written language is, in spite of our false sense of security in the rules of grammar. The ability of a text to be read by a machine does not confer infallibility and comprehension to that machine — on its own or as a conduit to a user. Despite Strunk & White’s efforts, there was a reason they wrote of elements of style.

*********************************

I find cataloging and related record creation to follow along these same lines. It too is a system of communicating information that is more of an art than its (multiple) rules based structures may suggest. It is also a system in which the art and stylistic traditions can and do fail machine readability, even when specifically designed for such.

You see, part of the art of cataloging is a series of visual tropes or signifiers that denote concepts or avenues of interpretation. For instance, [bracketed text]. This typically does not mean that there are actual brackets in whatever text, but rather may mean “This is an unofficial value I made the decision to assign to this field”…or maybe “I couldn’t read the handwriting and this is my guess”…or maybe “This is commentary, not actual information from the object”. Likewise brackets may be used to contain text such as […] or [?], which could mean “Unintelligible”, or “Maybe?”, or “Just guessin’ here”. (That last is my own way of sayin’ it.) One sees this frequently with titles, dates, and duration especially, though such visual signifiers may appear anywhere in a record.

And these markings are a thing of beauty — poetic layers of expression in six line strokes. A toehold of information in an uncertain world. A reflection of the cataloger’s fidelity to truth, even when it means admitting the truth is unknown and that one has failed in identifying it.

But.
But.
But, as at times can be the case, such art is impractical, limiting the efficiency of working with large data sets, or the advantage of machine readability to cleanup and search data, or the achievement of goals associated with minimal processing, which for archives is often the first pass at record creation.

As I have argued before, the outcomes of processing legacy audiovisual collections should support planning for reformatting and for collection management. This mean creating records at or near item level. This means data points like format, duration, creation date, physical characteristics, asset type, and content type must be quantified to some degrees. This means no uncertainties. There must be some kind of value for planning and analysis to work, whether it is an estimate or a guess or yeah-that’s-wrong-but-things-will-come-out-in-the-wash — or whether we just use common sense and context to support our assumptions rather than undermining them. This means be less precious.

The diacritics of uncertainty are meaningful, but when analyzing data sets they can push segments of records aside into other groupings when they likely shouldn’t be, and they make it difficult or impossible to properly tally data for planning and selection. A bracketed number or number with text in the same field is not a number to a machine. It cannot be summed. An irregular date cannot be easily parsed and grouped within a range. A name or term with a question mark is a different value than that same term with the question mark.

We know all of this — it is why we have controlled vocabularies and syntactical rules, and why we fret about selecting the “right” metadata schema. There is much theory around archival practice that comes down to exactness and authenticity of objects and documentation. However, there is a division in how such things are expressed and understood by a computer and by the human mind. There is also a division in exactness as data and exactness as a compulsive behavior. And any exactness is laid to waste by the reality of how collections are created.

When processing for reformatting, it must be kept in mind that the data created is not the record of record. It is a step to a complete record. It can be revised when content is watchable, or when embedded technical metadata gives us exact values like with duration. We know we are wrong or uncertain. The machine does not need to know it, cannot know it. But that fact will be apparent enough to future generations over time, as it is in so many other areas.

Joshua Ranger

Tags: , , , ,

Leave a Reply