Metadata schemas and tools to support digitization workflows
DWIM also means "Do What I Mean" -- this DWIM should try to do what the human operator would do by default.
Discarded name choices included: "deathworm", "deworm", "dewworm", and "dwalm".
This is really just a pie-in-the-sky proof of concept / thought experiment kind of thing. Its origins are likely related to a sleepless night after eating junk food right before I went to bed. Curse you, Voortman vanilla wafers!
We have four major areas where we digitize content:
- Digitization Services:
- Time-based media - uses the MDPI workflow tools
- Image-based media - uses ImageProc and varous scripts
- BDPL for born-digital content - uses manual processes and scripts
- IULMIA for film - this is a new workflow that is TBD
All of our digitization workflows follow the same overall structure:
- Someone requests some things to get digitized and delivers the physical media along with some documentation describing the metadata, etc.
- The digitization specialist will...
- prepare the media for digitization, which may include cleaning/repair
- digitize the media and verify it looks or sounds "right"
- process metadata is recorded (even if it is "this one is done")
- the files are pushed to a server somewhere
- return the materials to the requester
- Automated systems will...
- run some technical QC checks
- create derivatives (as needed)
- distribute the derivatives (as needed)
- store the data in a preservation / archive system
Right now the processes that we use for digitization, processing, distribution, packaging, and metadata collection vary greatly between different projects, media types, and sometimes within the same project. Most of this is because we've created one-off processes that were specific to the need at hand and they became the "go to" system for far longer than the project's lifespan.
Additionally, most of our workflows are centered around each physical object (PO) individually, rather than around the intellectual object (IO) as a whole.
Ideally we should have a human- and machine-readable record of how physical media became a digital object that can be stored with the object in a preservation system for future reference.
This primarily focused on these areas:
- Describing the structure of the physical media components comprising
the intellectual object
- order and description of the physical media
- purpose of the physical media ("performance program", and the like)
- Description of the physical media
- Physical description (type, size, length, media vendor, etc)
- Information required for successful playback (speed, groove size, etc)
- Conservation concerns
- Inspection details (warping, damage, fungus, etc)
- Conservation actions (baking, freezing, cleaning, etc)
- Digitization information
- Equipment setup details (software configuration, signal chains, etc)
- Basic information of the produced files (duration, color space, etc)
- Processing / Packaging / Distribution Information
- Provide hints to derivative creation (such as "audio is on stream 3")
- Override any automated QC checks for special cases
- defer distribution to an access/preservation system, etc.
- Processing logs, qc checks, etc will be collected
This information is explicitly not in scope except to support the functionality above:
- Descriptive metadata
- Only the bare minimum metadata is used to identify the physical media
- Identifiers can be included to refer to the system of record
- Ownership / Access control
- Only supporting communicating with the owner and returning media
- TBD: processing hints?
- Details about the content
- Only information relevant to creating/playing the digital copy is allowed, such as "audio volume is low on source" but not "speech begins at 3 minutes"
Details of the human workflow for digitization, such as project management or tools used, are not prescriptive -- they are intended to be illustrative of how in-scope functionality could fit into the bigger picture or to provide an example of how one could include this functionality.
The first pass of this concept is targeted directly at AV and Film digitization since there is a desire to retire the MDPI workflow tools and IULMIA will be adding a new Film scanner in the coming months.
The files used for storing metadata will be YAML. The decision to use this format is detailed in Why YAML
Where possible, the YAML files should be validated against a JSON schema for structural correctness and logically validated when processing.
Each IO will be given a organizationally unique identifier. These can be barcodes, VA numbers, etc. This ID will be used as the folder that contains all of the metadata, POs, and sidecar files that make up the IO.
NOTE: Any YAML examples here are based on the schema as it exists at the time this was written. The schemas will change as more details are fleshed out.
A request for digitization should include information about the intellectual object(s) that need to be digitized. In addition to basic metadata (such as title), these things should be present (if possible):
- Identifiers for the intellectual object, such as IUCAT barcode, catkey, call number, or a locally-scoped identifier (i.e. something off the spreadsheet)
- Information about the physical objects which make up the intellectual object
including indentification information (such as label text or barcode), and
an indication which order multiple objects should be presented. For example,
"the tape labeled '342' should be first, 'a13' should be second".
Additionally, if there are any missing items, they should be identified ("part 3 is missing") - The media to be digitized :)
The example will be a simple one:
- title: "A beginner's guide to potato hubandry", author: "t. uber"
- a single audiocassette with a case labeled "potato"
- call number SB211.P8 B56
NOTE: EVERYTHING BELOW HERE IS WOEFULLY OUT OF DATE
Every digitization project will need an ID -- it could be a VA number or some
other identifier. We'll use VZZ9999 as the ID.
The person doing the digitization will create a project directory in whatever
workspace is assigned. For an AV digitization, the space is on capybara and
the directory will be named the same as the identifier. So the directory is
also named VZZ9999
A skeleton digitization metadata file is copied into the workspace which a
name that includes the identifier. The skeleton would be saved as
VZZ9999-project.yaml and look something like this:
# yaml-language-server: $schema=schemas/core.json
version: dcs-digitization-workflow-metadata@v1
project_information:
id:
request_date:
completion_date:
contacts:
- name:
email:
notes: No Notes
descriptive_metadata:
title:
author:
identifiers:
-
constituents:
- label:
physical_media_description:
and filling out the bits from the project request:
# yaml-language-server: $schema=schemas/core.json
version: dcs-digitization-workflow-metadata@v1
project_information:
id: VZZ9999
request_date: 2024-07-16
completion_date:
contacts:
- name: Brian Wheeler
email: [email protected]
notes: No Notes
descriptive_metadata:
title: A beginner's guide to potato hubandry
author: t. tuber
identifiers:
- call_number: SB211.P8 B56
constituents:
- label: potato
physical_media_description:
The completion date remains empty and the label for the first constituent object has been set ("potato").
The physical media is an audiocassette, so moving the cursor below the physical_media_description field and indenting, CTRL-space lets us select the audiocassette information
Selecting it extends the file appropriately:
# yaml-language-server: $schema=schemas/core.json
version: dcs-digitization-workflow-metadata@v1
project_information:
id: VZZ9999
request_date: 2024-07-16
completion_date:
contacts:
- name: Brian Wheeler
email: [email protected]
notes: No Notes
descriptive_metadata:
title: A beginner's guide to potato hubandry
author: t. tuber
identifiers:
- call_number: SB211.P8 B56
constituents:
- label: potato
physical_media_description:
media_type: Audiocassette
position: 1
cassette_type:
tape_type:
tape_stock_brand:
format_duration:
conservation_concerns:
pack_deformation: None
damaged_tape: false
damaged_shell: false
fungus: false
soft_binder_syndrome: false
other_contaminants: false
remediation_notes: No Remediation
sequences:
- id:
tape_speed: "1.875ips"
noise_reduction: Unknown
sound_field:
duration:
notes: No Notes
There's only one tape, so the position '1' is correct. The tape overall tape information is noted:
physical_media_description:
media_type: Audiocassette
position: 1
cassette_type: Compact
tape_type: I
tape_stock_brand: Scotch
format_duration: 90 min
conservation_concerns:
pack_deformation: None
damaged_tape: false
damaged_shell: true
fungus: false
soft_binder_syndrome: false
other_contaminants: false
remediation_notes: Taped the shell together
The digitization for side 1 is performed using industry-standard tools and settings are noted:
- id: side 1
tape_speed: "1.875ips"
noise_reduction: Unknown
sound_field: stereo
duration: 22:30
notes: Recurring popping on the left channel
and the media file is named appropriately and put into the project directory.
Side 2 is handled similarly
The digitizer will do any visual/listening QC work to make sure the content is what is expected.
Now the digitization is complete and the "completion_date" field is filled out.
TBD:
- mapping objects and sequences to filenames, and filename conventions generally. Maybe the names are literally specified in the sequence itself?
- need some place to hang signal chain information
- need a place to put who did the digitization
When the digitization is finished the project directory is moved to a processing dropbox.
The processing system will look at the project file and:
- validate the project file, both against the schema and for consistency
- make sure all the files are there
- do any format-specific automatic QC for the file formats