Q1: Is primariness or secondariness (let's call it primacy) a property 
of a speaker, of the way a speaker is indicated, or of a speech?
A : you say that a character can be primary in one scene and secondary 
in another. So it is not a property of the speaker but of the speech. 
Does the way the speaker is indicated in the source indicate their 
primacy? My guess is not, so if you want to indicate it by an attribute, 
that attribute belongs to the <sp> element, not the <speaker>

Q2: Is primacy something you can automatically calculate (e.g. by word 
count, or by counting who else is on stage at the time)? If so it may 
not be necessary to encode it at all. I suspect this isn't the case 
however. If it involves human judgment and analysis,  you have the usual 
gallimaufry of TEI analytic mechanisms at your disposal: predefine 
possible values in a <category> and use @ana to point to them would be 
my choice but ymmv : you could add a child <desc> to the <sp> in your 
ODD and put all sorts of prose hoo hah into it. You could use (argh) 
@type on <sp> with or without a predefined list of values. You could 
invent your own attribute, as you suggest (but I wouldn't recommend 
@level for this purpose)

socratically yours


On 16/03/18 02:59, Elisa Beshero-Bondar wrote:
> Dear Stuart and Martin,
> We're somewhere in between both of your positions. Multiple characters 
> might be considered "primary", and multiple might be considered 
> "secondary" in a given TV episode. (When the team wishes to encode 
> that value only once it is for each character in the episode. They're 
> simply not using a castlist on each episode. Multiple characters might 
> be seen as equivalently "primary" as in, sharing the spotlight.) Could 
> this be calculated by string-length of speech? Sure, but this could 
> miss some characters in the animation who are important in a primary 
> way but aren't speaking much, or appearing only in the stage 
> directions. There are some complexities to the Rick and Morty show 
> that involve characters enhancing themselves across dimensions (in 
> which the screen is split, or so the students tell me) and the same 
> character speaks in different parallel universes in a state of divided 
> persona that they are deeming a third value of @level: "dimensional". 
> I'm not questioning it. But this seems useful data for them to be able 
> to extract, and I think the distinction between a dimensional vs. a 
> primary or secondary level may be a matter of human interpretation 
> rather than calculation in this instance.
> Of course I can suggest the string-length querying to them, since it 
> may be useful for them to determine who speaks the most, and which 
> characters turn up to be the most spoken about (a standard sort of DH 
> question for performance texts, narratives, and other sorts of 
> writings to do with the social interactions). Even so, they may want 
> to process the texts to make such a calculation and still output 
> something of the results of such queries in an attribute, and my 
> question stands, whether the adaptation of @level to this purpose 
> seems heretical, and whether there are perhaps better ways I may have 
> missed.
> It strikes me in writing this out that this sort of encoding is beyond 
> the pale of standard values and interchangeability, and that perhaps 
> defining a custom attribute may be the best path.
> Cheers,
> Elisa
> On Thu, Mar 15, 2018 at 10:48 PM, Martin Mueller 
> <[log in to unmask] 
> <mailto:[log in to unmask]>> wrote:
>     That's been my experience with capturing the number of speeches
>     and spoken words in several hundred early modern plays. It's a
>     very primitive measure, but it's easy to calculate, and it works
>     surprisingly well.. But if the resulting numbers are very close,
>     "primary" and "secondary" cease to be useful categories.
>     *From: *"Stuart A. Yeates" <[log in to unmask]
>     <mailto:[log in to unmask]>>
>     *Date: *Thursday, March 15, 2018 at 9:42 PM
>     *To: *Martin Mueller <[log in to unmask]
>     <mailto:[log in to unmask]>>
>     *Cc: *"[log in to unmask] <mailto:[log in to unmask]>"
>     <[log in to unmask] <mailto:[log in to unmask]>>
>     *Subject: *Re: Primary vs. secondary characters
>     Does this actually need to be encoded manually? If you're already
>     encoding the speaker of each word/<p>, you can go though after the
>     fact and calculate how many words each person said in each
>     episode. From that you should be able to calculate the relative
>     prominence of each relatively straight-forwardly...
>     cheers
>     stuart
> -- 
> Elisa Beshero-Bondar, PhD
> Director, Center for the Digital Text | Associate Professor of English
> University of Pittsburgh at Greensburg | Humanities Division
> 150 Finoli Drive
> Greensburg, PA  15601  USA
> E-mail: [log in to unmask] <mailto:[log in to unmask]>
> Development site: <>