Timecode Metadata

Timecode metadata are the critical link in the between textual content and audio or video in digital environments. Different architectures for timecode deployment have evolved independently in the creation of digital oral history collections, and all help to significantly increase digital accessibility. With many models now on the table it is an appropriate time to take inventory of what approaches are available, closely evaluate the relationship between these models, understand the range or textual data they are linked to, and elucidate the current “state of the art” to find common ground for future developments. 

Timecodes are being put to use in two broad ways: 1.) as transcription timecodes, enhancing full text transcriptions with a cross-reference to time points in the source audio or video, and 2.) as audio or video file metadata enhancing a longer audio or video file, or A/V timecodes. Within A/V timecodes two basic models are emerging, one that uses timecodes pointing to a single point in time in the digital file, allowing the user to play forward from that point. (We might call these indexing point timecodes.)  In another model, (which we might call passage timecodes), timecodes are defined as inpoints and outpoints giving meaningful content within a longer digital file its own begining, middle and ending. 

The latter model of defining passage timecodes can take place in database environments where the in/out points are just references that move the listener digitally (hypertextually) to the passage of interest. In other contexts, practitioners manage oral histories by hard-editing passages permanently, thus creating segments or clips from the full length digital source file.   

All timecode deployments require choices to be made--regarding the frequency of transcription or indexing point timecodes, or the length and comprehensiveness of passages timecodes. No standards have been set as to how these choices are made and there are strengths and weaknesses of the different approaches. I hope to have the opportunity to compare notes with others using the various models, determine the trades-offs between models, establish what can and cannot be standardized, and allow digital oral history stewards to proceed with future investments in software more informed.

 

Representation of recordings through Annotation

Our practice of Oral History content management, which we often refer to as digital indexing, began by questioning the assumption that recordings must be transcribed word for word before they can be used.  In the database-driven environments we work in, summary annotations are much preferred to full transcription. The work of the annotator is at the heart of digital indexing and we continue to reiterate that full transcription is always an option if the need is there and time and resources allow. Here are some random musings about annotation...

• Annotation is about representing what is on a recording in a text format.  In this sense, it is no different than a transcription.
• An annotation needs to describe passages of audio adequately enough to lead a user to that passage. Defining the users well may be as important or more important than the specificity with which the annotation represents the passage.
• An annotation can be enhanced by using strategic vocabulary words within the prose of the annotation. Thus full text searches will get hits on that digital object (passage of audio or video). 
• All annotations are subjective, and that is totally okay. It is all part of recognizing, defining, and composing toward an audience of users. Our subjectivity saves them time.

Annotator

Annotation is a key step upon which indexing is built, and the role of the annotator is to link the content of the audio or video with meaningful text. Annotation is itself a form of indexing—creating text that directs a user to content of interest—that simply takes a linear form, similar to a transcript. The basic skill is listening to a recording, and composing text that summarizes the content. Although annotation seeks to lighten the tedious burden of word-for-word transcription, it still takes significant time to complete (at least 1 ¼ hours per hour of interview, sometimes up to 2+ hours or more if a great deal of detail is desired). Annotation proceeds faster with practice and with increasing familiarity with the content. Familiarity with the content from the outset (whether because of personal background/interest or because the annotator conducted the interviews) is typically an additional advantage. Because annotation is the foundation upon which the indexing is built, consistency of style, density, and overall quality of annotation is strongly recommended. Generally, annotations that are shorter but consistent are better than a highly varied collection of sparse and detailed annotation written by different people in different styles. When there is more than one annotator, an editor is an essential player in the indexing process. 

Return to Oral History Digital Indexing Roles.

Starting with Indexed Summaries

The assumption that working with oral histories requires transcription first is one we have been examining for years through our practice of digital indexing. Transcriptions can be a powerful resource to accompany recordings, but we choose to approach oral history analysis differently.

With the availability of database-driven models for oral history content management, we start with summary annotations and create indexes, and work from there. Selective or full transcription is always an option, and in the meantime, summary-passage architecture for content management is an efficient and functional way to engage oral history recordings. In addition, the resulting text can be less dense than a transcription, more contextually meaningful, and still be suitable for all sorts of publications and for further development.

In a nutshell, we respond to the “unexamined assumption” with another question:

Does it make sense to transcribe everything you have so you can find the few passages you really want? Or does it make more sense to index a collection first and then transcribe what you need when you need it?

For us, the latter makes more sense and this is where digital indexing begins.