Work to date:
Lee tested OHMS for indexing and syncing the transcript and audio of a single 30-minute interview currently in the Digital Collection (Ronald Duarte). Multiple attempts to sync the transcript and get it to display well gave rise to the following lessons about transcript preparation before upload to OHMS:
Remove front matter and the date/session headers
Strip out any hidden codes (used for indexing in some transcripts)—single-step search and replace process.
Remove triple hard returns throughout transcript and double spaces after periods—(simple search and replace processes).
After upload, add one hard return to the start of the transcript for the best display results, then sync.
It’s possible to edit the uploaded transcript, but it’s best to keep it to minimum. Any edits to the uploaded transcript which affect the line spacing of sections that have already been synced will throw the sync off.
Lee experimented with creating segment-level indexes for this single transcript at two different levels of OHMS indexing (see OHMS Indexing Levels).The first segment is indexed at Level 3; subsequent segments just have the segment title and synopsis (copied from the transcript’s table of contents), along with a partial transcript. This puts them a bit ahead of level 1, but for level 2 they would need some keywords. Next steps will include adding keywords to these segments and trying to index and edit a longer interview.
Questions / Issues for discussion:
What item-level metadata fields should we be filling in to maximize utility and minimize duplication of data with the Digital Collections?
Title (required by OHMS)
Media Format (required by OHMS)
Accession number (use OH/RI number; decide on format)
Series ID and Name (use to categorize interviews by Project; useful on OHMS backend for workflow management; front-end can link to COH website project info)
User Notes: “can be used for notes that an institution wants communicated to the public user each time the viewer is opened. The message can be closed by clicking on the X in the right corner. The User Note will reappear each time the public user switches between the transcript and the index.” (use to help clear up possible confusion about edited transcript vs. unedited audio / multi files vs. single file / how to cite?)
Acknowledgement (could use language from project/sponsor acknowledgement page)
Possibility for batch import? See OHMS_Import_Template
For help thinking this through, see: OHMS Metadata Fields
Final decision to be made in consultation with DigColl staff.
How are we preparing audio/video files for upload?
Stitched together submasters/use copies for legacy interviews already in DigColl—to be handled programmatically by developers
Stitched together submasters/use copies for legacy interviews not yet in DigColl--??
Revise OH workflow going forward to create stitched together submasters as basis for transcription. Update “Note on Transcript” in front matter to indicate change.
Chuck on board for figuring out and dealing with storage/preservation implications.
How do we want to handle OHMS indexing (segment-level metadata)?
What level of depth do we want to go into? See OHMS Indexing Levels
Whatever level of indexing depth we choose, this is the work- and time-intensive part of the process for COH staff.
Indexing at least to level 2 is recommended if we want to justify no longer creating back-of-the-book indexes on the basis of the discoverability/searchability boost we’re getting from implementing OHMS.
Existing table of contents gives us starting point for Title and either Keywords or Synopsis fields (both?). Simple cut-and-paste plus transcript/audio navigation to tag segment beginnings gets us to level 1 at a pretty low effort.
Level 2 adds partial transcript (low effort), either keywords (low to medium effort) or subjects (high effort) in addition to synopsis. Segment breaks at 4-6 minute intervals—probably corresponds roughly to our existing tables of contents, but may be a little more frequent. If we eschew subject headings entirely, we can pretty easily handle level 2 indexing. In 2/18 DigColl Tech check-in all parties agreed that forgoing segment-level subheadings makes sense for the foreseeable future.
Level 3 calls for a more detailed synopsis, both keywords and subjects, and the inclusion of external hyperlinks and GPS coordinates when appropriate. Segment breaks roughly every 3-6 minutes. Level 3 is aspirational at this point. We might try it out for particularly popular/interesting interviews.
Do we plan to upload transcripts and sync them with the audio files?
How much work is involved in prepping a (legacy) transcript for upload? (see above work-to-date note for lessons learned on pilot interview transcript)
When the published transcript is relatively faithful to the audio (very little)
When the published transcript deviates significantly from the audio (it depends)
Attempt to sync a heavily edited transcript with audio that it diverges quite widely from? (do not recommend—DJC agrees)
Return to an earlier, RAW or A/E version of the transcript for use with OHMS? (recommend)
Need to be aware of any redactions that were made post-AE and must be preserved for legal reasons, and edit audio file to match, if necessary.
Create a new, verbatim transcript and audit it for accuracy, then upload? (only if appropriate earlier transcript version is not available)
Again, need to be aware of any redactions that must be preserved for legal reasons, and edit audio file to match, if necessary.
Subcase: video, esp. with flawed ASL interpretation
What time intervals do we want to use for the sync feature?
One minute appears optimal, and has DJC approval.
Revise transcription order to request time stamps at OHMS intervals, formatted correctly for upload. This will save us having to manually sync in the future.
Revise workflow to preserve a stripped-down audit-edited transcript as the OHMS version (this one will lack hidden codes, front matter, index, session headers and extra spacing, see above). Make sure that any permanent redactions made at a later editing stage are incorporated into the OHMS version and into the version of the audio that will be made available to patrons (replace redacted audio with a tone that runs for the duration of the redaction).
Consider revisions to release forms to more explicitly state our intentions with regards to making less-heavily-edited versions of the transcript available online as a transcript that accompanies the audio, either with interviewee permission or after their death.
What will we offer patrons the ability to download via the Digital Collection app?
Multipart audio files
Edited/published version of the transcript PDF, corresponding to multipart audio files but also possibly deviating from the audio considerably
Single, stitched together audio file
Transcript corresponding (both in time coding and in substance) to single, stitched together audio. This can be autogenerated as a PDF file via the OHMS viewer (including in preview mode). See Interview-export1581704695.pdf How customizable is this feature?
The developers want to create a customized viewer for deployment on the DigColl app, rather than relying on the default OHMS viewer. If we insist on the need for the viewer to include this PDF generator feature it’s more than likely that they’ll abandon plans for a custom viewer and use the standard one instead. However, we could generate the PDF using the OHMS preview feature from the backend and then manually upload that transcript to the Digital Collection ourselves as another derivative product that we make available.
Decision to be made primarily by development team, in consultation with COH.
How do we help future patrons figure out how to use/cite the different derivative products of each interview?
User note field (see above under item-level metadata fields)?
Other?
We have “how to cite this interview” info in each published transcript.
If the published transcript adheres closely to the audio, patrons can use that citation format and be confident that future researchers will be able to verify the citation, whether the patron relied on the published transcript, the audio, or the OHMS sync’d transcript.
If we’re doing our best to sync a heavily edited/non-verbatim transcript with the audio, then we need to inform the patron that the audio and transcript diverge significantly and instruct them on how to cite the audio specifically (important if they’re pulling a clip/quote from the audio that doesn’t match up with the transcript). [This whole approach seems unworkable. DJC agrees. We’ll omit it from further consideration unless circumstances arise that require us to put it back in the mix.]
If we’re using/creating a more verbatim version of the transcript for use in OHMS in cases where the published transcript deviates significantly from the audio, we need to inform patrons about the deviation and instruct them how to cite the audio/verbatim transcript to differentiate it from the published transcript. This information should be displayed with/by the OHMS viewer (like the User Note field) and should be included in a PDF version of the verbatim transcript (the User Note field is).
BIGGER PICTURE ITEMS
What records could we add to the Digital Collection/OHMS right now?
Completed research interviews where the Research Interview Release Form was signed. (CUSP, others?)
Drosophila interviews? (Not sure—think we need separate internet release forms for those). Not transcribed. COH staff will review these and determine after the OHMS feature is implemented.
Women engineers with release forms? Not transcribed. COH staff will review these and determine after the OHMS feature is implemented.
How can we make additional interviews available via DigColl?
For free access interviews without internet permission, post front matter/index files and accompanying metadata; no audio, no full transcript. DJC approves.
Ask DigColl team about feasibility.
Begin identifying interviewees to ask for internet posting permission. (Wait until OHMS launched for DigColl so we have something to show them). DJC approves.
Where does the supplemental metadata we have on the COH microsite really belong? This is the metadata about the INTERVIEWEE rather than about the interview—birthdate and birthplace, deathdate and deathplace, education, professional experience, and honors.