Toward a Blackbird Style Guide for Captioning Poetry

by M.A. Keller

An example of my preliminary questions about captioning poetry can be found in my previous Birdlab blog entry “Searching for Captioning Best Practice—Poetry.” The post provides just a few examples of problems the journal staff faced while trying to create and refine captions for a reading of untitled short poems by Ellen Bryant Voigt published in v17n2 of Blackbird.

Colleagues who do not work in production have an overly optimistic view that “we’ll just look at what someone else is doing” to find a solution; then the publications/organizations offered as sources are not doing any captioning of video at all, so where to start? The Poetry Foundation has one article referencing closed captioning “Poets, Turn on Your Close [sic] Captions,” but it centers not on delivery of poetry to deaf and hearing impaired communities, but how (presumably hearing) writers can use the poor quality of closed captioning in general to find idea for poems through error and ridiculous juxtapositions created by mechanical (auto) captioning.

“Should poems and other quoted material be captioned as they were originally written?” asked Sean Zdenek in a 2011 blog post titled “Iambic Pentameter Captions?”  His answer to that (in short, “yes,” but be sure to read Zdenek’s entire post) aligns squarely my thinking, even as Zdenek points out limitations of line length in differing captioning formats. Zdenek is an associate professor at University of Delaware whose work centers on captioning and disability studies; I have previously distributed to student editors the first chapter of his Reading Sounds: Closed-Captions and Popular Culture (University of Chicago Press, 2015).

I have just recently discovered, as a base stylebook for captioning,The Captioning Key, a resource provided by the Described and Captioned Media Program (DCMP) as “funded by the U.S. Department of Education and administered by the National Association of the Deaf.”

Blackbird Production Goals for Closed-Captioning Poetry

1. Use The Captioning Key as a base stylebook, and Zdenek’s call for an awareness of the poetic line in captioning (within character-display constraints), to develop a guide to best practice for captioning poetry.
2. Determine if Blackbird student editors and staff can readily execute these guidelines with current tools.
3. If primary VCU tool (Kaltura) is limited or difficult, search for a solution in our existing software library and/or look for another software solution
4. In discussion and planning, keep both product and process in mind; this isn’t just about producing “good enough” ADA compliant text; it is about teaching student interns and editors how to produce a product that best serves the audience and the genre. To that end, outsourcing undercuts the journal’s educational mission as well as our possible contribution toward developing best practice.
5. Use “A Reading from Kyrie” in v17n2 as a test case and proof-of-concept.

DCMP’s The Captioning Key opens with the logic of breaking text for readability under the section “Line Division,” stating “When a sentence is broken into two or more lines of captions, it should be broken at a logical point where speech normally pauses.” A series of clear examples follows, suggesting that captioning tools should readily allow control of the breaks and arrangement of text in any given caption window.

Poetry, however, complicates “the logical point[s] where speech normally pauses.” Because of this, the logic the The Captioning Key applies to prose may not be the best practice for poetry. What is the responsibility of a captioner to privilege the boundary of the line over the sentence when captioning poetry?

As I work on preliminary questions, at first solo, and then later with student editors, I will append my progress below in this document to keep discovery adjacent to the initial questions. From there, I’ll decide if something has the weight to warrant further explanation in a separate post.

Additional reading

Interview with Adam Pottle, author of Voice: On Writing with Deafness
Ableism, captioning, deaf culture, writing, etc., February 27, 2019

Designing Captions: Disruptive experiments with typography, color, icons, and effects
Sean Zdenek’s web text on experimental captioning in Kairos 23.1, Fall 2018

Searching for Captioning Best Practice—Poetry

I am looking for guidance for online literary publishers in captioning poetry for audio and video presentations. We want to get this right and/or to help develop best practices. Questions of line and stanza break come in to play with poetry, as does where one poem in a series ends and another begins, say, if a writer is reading from untitled works (as was the case in our first attempt at captioning). What follows is a description of attempts, problems, and questions coming out of our attempt, and failure, to get this right using Kaltura, the one tool currently available to us. It is a question of control and the ease with which we can exercise control (and train interns to do so) with the available tool.

What is described below is an attempt to capture what we observed under deadline and duress, and hopefully I will find a ready solution, but in the meantime, I want to go ahead and blog a version of the notes and observations lest anyone else have an obvious solution or can direct us to a captioning tool that makes some of this work less difficult than in Kaltura.

—M.A. Keller

I want to begin by giving the whole of two poems as they would appear on the page, one poem per page, and that were read back to back by Ellen Bryant Voigt in a video Blackbird published in v18n1. I will then follow with how the first of these texts appears in the captioning, and how the em dash, at the end of the first poem, makes it appear as if it, and the poem that follows, are a single unit. (FYI, we submitted a mechanical caption request to Kalturea and then corrected the text afterwards.) For context, the poems come from Voigt’s book Kyrie, a sonnet series of voices speaking out of the 1918 influenza pandemic. The first poem example here in the voice of a young speaker moving into delerium; the second in the voice of a teacher.

My brothers had it, my sister parceled out
among the relatives. I had it exiled
in the attic room. Each afternoon
Grandfather came to the top stair, said
“How’s my chickadee,” and left me sweet
cream still in the crank. I couldn’t eat it
but I hugged the sweaty bucket, I put
the chilled metal paddle against my tongue,
I swam in the quarry, into a nest of ropes,
they wrapped my chest, they kissed the soles of my feet
but not with kisses. Another time: a man
stooped in the open door with her packed valise,
my mother smoothing on eight button gloves,
handing me a tooth, sprig of rue—


All day, one room: me, and the cherubim
with their wet kisses. Without quarantines,
who knew what was happening at home—
was someone put to bed, had someone died?
The paper said how dangerous, they coughed
and snuffed in their double desks, facing me—
they sneezed and spit on books we passed around
and on the boots I tied, retied, barely
out of school myself, Price at the front—
they smeared their lunch, they had no handkerchiefs,
no fresh water to wash my hands—when the youngest
started to cry, flushed and scared,
I just couldn’t touch her, I let her cry.
Their teacher, and I let them cry.

Although these are two separate poems, and although there is a twelve second gap of silence between the ending word “rue” and the beginning phrase, “All day,”  of the next poem in the video, Kaltura’s mechanical caption placed “rue—” ending the one poem on the same caption display as part of the first line of the next poem.

In another attempt, I replaced, though incorrectly quoting the poem, the em dash with a period after “rue”  Then we got this, which is a different problem

Individuation of poems aside, we first considered indicating line breaks and stanza breaks with single and double slashes respectively, as per print convention in quoted text, but after I tried this for the entire video, I found several significant problems illustrated in the following notes and screen captures from the beginning of the reading, quoting the following lines (only part of the poem displayed below).

Dear Mattie, You’re sweet to write me every day.
The train was not so bad, I found a seat,
watched the landscape flatten until dark . . .

  1. Kaltura did not retain a slash on the same display screen by default at the end of the line following “every day”  but confusingly began the next caption screen with the slash.

2. When I cut down the sound and just read the captions, the slashes too often read as the italicized capital letter “I,” as in the example above, throwing the text into confusion.

3. Kaltura’s captions are not left bound, so we get both Kaltura-created caption display breaks and centered text as in

Dear Mattie,
You’re sweet to write to me everyday.

Given the overly common problem of poems being amateurishly reset on a center default, this looks, to a copyeditor/typesetter/pagebuilder, naively deliberate and incorrect, and suggests wrongly that line breaks are being applied in the caption display.

So our questions:

  1. Can we readily display one line of poetry per caption screen (even if it wraps to two caption lines in the caption display space.) Even if we don’t really intend to overtly signal line or stanza breaks (as with slashes) working with the line as a unit makes more sense in the captioning than the default of Kaltura breaking captions based on punctuation, especially a poem may be lacking punctuation, or using it in an atypical fashion, such as an em dash at the end of a poem instead of end stop punctuation.
  2. If we opt for / and // for line and stanza, can we exercise control over their placement on the caption display.
  3. In either case, how can we clearly signal in the caption display the end of one text, and the beginning of another, especially in untitled sequences.
  4. Can we easily eliminate the center default?

Resources for Research & Research Editing

The following list includes resources I demonstrated to the 
Blackbird copyediting team, Feb 1, 2019. —mkeller

Texts and Image Archives

Google Books

Internet Archive

Galica (Bibliothèque nationale de France)

The New York Public Library Digital Collections

Folger Digital Image Collection
Recommended and demonstrated by Chris Alimenti

Newspaper Databases

Virginia Chronicle (Library of Virginia)

Chronicling America (Library of Congress) /
Subscription only. On-site access at no charge at The Library of Virginia

Genealogy Resources

Find A Grave

Ellis Island Passenger Search

Family Search
Church of Latter-Day Saints. Required registration for free account to access databases.


The End of Hot Type

The 1978 documentary linked below is one that we show to interns every semester as we talk about the history of publishing and related technologies.  We highly recommend it.

Filmed on July 1, 1978, this documentary by David Loeb Weiss chronicles the end of “hot type” at The New York Times — and the introduction of computers into The Times’s printing process.

Farewell Etaoin Shrdlu

Legacy Media, Migration, Mechanical Captioning, & Migraines

[Text below drafted, but not published, early summer 2016, as I was trying, solo, to make a decision about whether or not to migrate Blackbird’s legacy media files to Kaltura, a video system adopted by VCU.  Ultimately, I decide not to use that platform for the legacy files and in the first pass at mending he archive last summer, I have converted much of the archived media into mp3 and mp4 files that are saved in the structure of the journal on the web server, rather than being hosted, as before, on a media server. —Michael Keller]

In the spring of 2016, VCU announced it would be taking down their “video server” on June 30, 2016. This server contained media files for Blackbird, vol 1-8 (legacy real media) and vol 9-12 (flash server). There were no plans to automatically migrate any of the legacy video server files to Kaltura, the new system; that responsibility rested with the individual faculty member, unit, or organization. This, based on the assumption that anything on the old video server was likely outdated, poor quality, or otherwise not of interest to migrate.

At the time of the announcement, Blackbird was working hard on the spring issue, and the magazine was down two expert staff members, having recently lost a faculty-line online editor and a graduate student lead developer. With this staff loss, I could not stop to investigate all the shutdown was going to mean for us come summer, though I  did email local Kaltura support queries about the possibility of a batch move of data to Kaltura, but my question was lost in the shuffle.

By the time we published the spring issue in May of 2016, and I was able to meet face-to-face with a very amiable support staff member about Kaltura, the problem was this: almost every question I asked in the meeting had not been raised by anyone before in the planning and execution of the server shutdown.

Blackbird‘s needs and problems were (as usual) unique to the journal and its history. The idea that we were trying to preserve an active fifteen year archive of material came as a surprise to IT staff members.

Although there were a number of issues, my chief concerns, after some self-study of Kaltura, were as follows:

1) is batch migration and conversion of 502 real media files from v1-v8 possible? What about flash server materials from vols v9-v12?
2) can the journal opt out of auto-captioning by Kaltura on upload since we cannot, of course, publish unedited captions.

At the time of my meeting with support, there was no batch import solution, which would mean that I would have to manually upload files one at a time to Kaltura, where they would be auto-captioned by default (at that time) in our institutional setup.

It turned out that

1) no batch option for file conversion was available
2) there was no option to opt out of caption and no way to hide the caption toggle in the player; the only option was to manually and individually delete captioning after the fact.

In addition, each manually converted file would have to be embedded on the page in a Kaltura player, which was optimized for video only; there is not really an “audio playbar.” Attempts to make one by reducing (and hiding) the video playback space, as we do with JWplayer, were stymied by the design of the playback controls.

On examining this possible solution, I realized the high number of times each of the v1-v8 502 files would have to be handled. 1) dowload from video server 2) rename* 3)import to Kaltura  4) tag (multiple tags, including volume and issue, genre, and contributor name) and wait for enforced, but unusable, captioning to complete 5) export captions for future reattachment and editing 6) rename exported caption files to match the name of the associated media file 7) copy embed playback code and file path. 8) place embed code on web page while dealing with any formatting/layout changes in legacy pages 9) publish to test server and check playback 10. have proofers/copyeditors do a thorough review of any formatting and/or textual changes and noting any changes to the document using code comments 11) republish to live server

So without taking into account other page issues (such as the cleanup of old code and deletion of local real media files containing paths that redirected to the video server (the way RealMedia worked), you have 11 actions (really more given the tagging, etc.) x 502 for over 5522 actions. That seemed a little much, especially if Kaltura is not an archival solution in and of itself, but a delivery system that would, itself, be replaced in some future update cycle.

A further note on captioning
Unedited captions were unpublishable. That’s a given. The examples below come from my first two tests of some departmental legacy audio.

In a Fizgerald lecture: “For his part Fitzgerald is going on the rag and often simply meant that in fact he would sway a awful fart liquor but would continue to consume vast quantities of beer and wine . . .”

In a lecture on the Waste Land, the poem title is auto captioned  as “The Worst Run” and “The Waist Line.”

But knowing that we would, at a later date, want to be able to edit, and make public, captions, I hoped we could, in the embedded player, toggle off the choice of captions as publishers; that is, we wanted to hide access to the captions until they were edited. This was not an option, so my only choice, after waiting for up uploaded file to be processed through auto-captioning, would then be to delete the captioning.

Well, I thought, if the captions will already have been processed, we should probably plan to save the caption file down until such time as we can reattach it and edit.

Turns out that when you export a caption file in Kaltura, it does not name it in alignment with the source file; in other words levine.rm does not save out as levine.DXRP. It saves out as english.DXRP, as in “English language,” as does the next, and the next. so you end up with a list of exported files that looks like


So if I wanted to retain the captions for further edits and reattachment, I would also have to stop to manually, and very carefully, save the exported caption file to match the name to its associated media file.

* [On renaming] Because the legacy source files came from a file and folder system, the file path, not necessarily the file name, identified the issue of the journal and the author of the piece. For instance, I only know that “interview.rm” below is a conversation with Phil Levine and in Blackbird v1n1 because of the containing folders.

Whether we has used Kaltura or not, it became clear that each file, upon conversion to a new format, would have to decide on a model for a more self-descriptive filename. Below is what I adopted on the fly, though I wonder/worry that we should have written out “blackbird” for “bb.”





Em Dashes Break “Smart Quotation” Algorithms


Quotation replacement algorithms routinely swap unidirectional quotation marks with directional quotation marks. (Examples below show unidirectional vs directional quotation marks in Verdana and Georgia typefaces.)


The majority of these algorithms do not, however, properly handle adjacent em dashes, en dashes, and hyphens, resulting in substitution of directional quotation marks that point in the wrong direction.

The examples below use em dashes to illustrate the problem. The following behavior is observed in Microsoft Word 2013, Windows, when the “smart quotes” function is switched on.


It seems that an em dash, en dash, or a hyphen preceding a quote are the instances in which a visible character (as opposed to a space or line break) will cause Microsoft Word to convert to a opening quotation mark; all other characters will produce a closing quotation.

Google Docs and WordPress share a variant of the problem, distinct from the behavior in Microsoft Word. (Most contributor manuscripts come to us in MS Word.)


Implemented practice

Blackbird’s copyeditors will be trained to watch for these problems in manuscripts.

Our internal character conversion mechanism has been updated to alert the pagebuilders to any instance of adjacent dashes and hyphens for an additional proofing.

Modifying a smart quote algorithm to be contextual for this extremely specific case seems to be a daunting task.  Any potential developer may consider using regex to check for the direction of the previous quotation mark solely.

Discovery, research, & TEXT by Joe Woods
Image Captures & Editing by M.A. Keller


Directional Quotation Marks, Primes, & Unidirectional Quotation Marks


This opening guidance on setting quotation marks comes from the Chicago Manual of Style 16, which uses four different terms for the preferred character pair at the outset.

6.112 Typographer’s or “smart” quotation marks

Published works should use directional (or “smart”) quotation marks, sometimes called typographer’s or “curly” quotation marks. These marks, which are available in any modern word processor, generally match the surrounding typeface. For a variety of reasons, including the limitations of typewriter-based keyboards and of certain software programs, these marks are often rendered incorrectly. Care must be taken that the proper markleft or right, as the case may behas been used in each instance.

This post privileges the terms “directional” and “unidirectional” as we catalog variants.

Directional quotation marks—alternately known as typographer’s, typographic, typeset, curved, curly, or smart.

Unidirectional quotation marks; alternately known as neutral, vertical, straight, typewriter, ambidextrous, programmer’s, or dumb.

(If there are alternates that we have missed, please let us know.)

In working parlance at Blackbird, we are more likely to say “typographical” and “typewriter” rather than “directional” and “unidirectional.”  I had come to prefer “curved” vs. “straight” because of the clear visual that language evokes. That said, this BirdLab post is meant to help us determine what language to use in our own style book, a decision not helped by Chicago’s choice to head their passage “typographer’s” and then to use the term “directional” in the opening sentence below the header.

I am decided on one point: quotation marks are neither “smart” nor “dumb,”  though these are popularly usedif misleadingterms. More on this to follow.

Here are four examples of opening and closing directional quotation marks in serif and sans serif fonts, including Blackbird‘s preferred body font, Verdana. (Note that the right single quote is the same as an apostrophe and is not otherwise discussed in this post.)


fig 1. directional quotation marks, double and single

For purposes of disambiguation, here are prime marks for these same four fonts. The following passage is quoted from the Wikipedia article on prime marks, which provides more complete coverage for pagebuilders than any mention of primes in Chicago.

The prime symbol ( ′ ), double prime symbol ( ″ ), and triple prime symbol ( ‴ ), etc., are used to designate several different units and for various other purposes in mathematics, the sciences, linguistics and music. The prime symbol should not be confused with the apostrophe, single quotation mark, acute accent, or grave accent; the double prime symbol should not be confused with the double quotation mark, the ditto mark, or the letter double apostrophe. The prime symbol is very similar to the Hebrew geresh, but in modern fonts the geresh is designed to be aligned with the Hebrew letters and the prime symbol not, so they should not be interchanged.


fig 2. prime marks

The primes are given here for comparison to the unidirectional quotation marks (fig 3) inherited from typewriter technology. Generally, we would call them “straight” or “typewriter” quotes but the term “ambidextrous quotes,” is a fitting historical descriptor as these figures are used on the typewriter keyboard to stand in for either prime marks or quotation marks. (“Ambidextrous quotation marks” is a term from Wikipedia’s article on quotation marks; I’m searching for uses of this language in other sources to see if it comes from typesetters or programmers.)

Unidirectional quotation marks (standing in for directional quotation marks or primes) should never appear on a page of Blackbird text; convert to the proper character in context.


fig 3. unidirectional quotation marks

David Dunham cites himself in an online page as the first person to write code to swap unidirectional quotation marks for directional quotation marks as early as 1986. Dunham writes

“Smart Quotes” is the automatic replacement of the correct typographic quote character (‘ or ’ and “ or ”) as you type (' and "). It does not refer to the curved quotes themselves.

Dunham confirms what I’ve been teaching for years: “smart quotes” refers to the automated conversion process, not to a set of characters themselves. This popular misuse has lead to the reverse labeling of unidirectional quotation marks as “dumb quotes,” but not in our shop.

When copyeditors encounter a contributor manuscript with unidirectional quotation marks, they note it on a paper copy. When the manuscript copyedits from multiple editors are reconciled to a digital final form for pagebuilders, the reconciling copyeditor applies the Microsoft Word “smart quotes” macro to the document to properly format quotation marks and apostrophes.

At the next step of copyflow, when the document lands in the digital “final for pages,” folder, pagebuilders must again revisit quotation marks upon converting them to HMTL. Information on this process will appear in an upcoming post

(For those with access to the online Chicago Manual of Style, read more at Special Characters.)

Also be aware that some systems, like WordPress (on which this blog is published) apply a “smart quote” filter to convert unidirectional quotation marks to directional; in draft, I see unidirectional quotation marks; on publication they are converted to directional quotation marks.

In this environment, if I actually want to display a unidirectional quotation mark, as in one instance of the Dunham quote above, I have to overtly mark up the unidirectional characters to keep them from auto converting. Be aware that this kind of auto conversion may muddle online discussions of this issue if an example of a unidirectional quotation mark is rendered, inadvertently, as directional. (I’ve seen this very problem more than once.)

<code>'</code> and <code>"</code>



Blackbird is careful to always set directional quotation marks, but we need to agree on preferred language for in-house style and production manuals.

M.A. Keller