Semantics of \frametitle

[josephwright]

In beamer, one can use \frametitle independent of \section, etc. Sections are printing out by some themes but not by others, whereas frame titles are (almost) always included in the typeset material. For example, the basic document

\documentclass{beamer}
\begin{document}
\section{Foo}
\subsection{bar}
\begin{frame}
  \frametitle{Some stuff}
  Hello world
\end{frame}
\end{document}

prints the frame title but not the section, while adding \useoutertheme{infolines} will make the section and subsection names appear on the slides.

When you look at this from a tagging perspective, this presents a challenge. We need to be able to give some form of structure, and it's clear that in most (all?) cases, frame titles are part of that. So the question is to whether frame titles fit into the structure of sections, whether they are parallel, etc.

One approach would be to say that frame titles are the level below whatever sectionning is currently applied, so for example if the user has a \section, \frametitle would be a \subsection in structural terms. But that then requires that the section appears somewhere in the output. One could imagine requiring that if sections are used, they appear, for example - that might be one solution.

Looking at PowerPoint as the most common GUI tool for making slides, it offers sections and frame titles in what does seem to be an independent way: sections are about the outline structure, frame titles are about the text. There's only one level of outline, but it does seem to appear in the PDF output if one exports - though I'm not seeing it as part of the tagging structure. So I am not sure this really gives us much of a guide: certainly it's a very different model to the beamer one.

Thoughts in this area are very welcome: it's not at all clear to me what is 'right' here. But I think we do need a consistent model - leaving this to individual themes (depending on whether they include sections) is likely not a good idea.

[wspr]

Sorry I’m not up to speed with the tagging details — is it truly impossible to have an “invisible” section from a tagging perspective? Assuming not. Personally it makes sense to me that you might require a “section slide” if you use sections at all — the whole point to a section is to break up a presentation and having a breather slide to do this makes complete sense (to me). Although I don’t know if there are all sorts of document constructions in the wild that don’t follow this (e.g., the section slides are omitted in one or other of the document output forms). I would say, conversely, if there is no section slide then you might omit that level of section tags. But I’m not sure if you can be that “dynamic” in your tagging decisions.

[josephwright]

Sorry I’m not up to speed with the tagging details — is it truly impossible to have an “invisible” section from a tagging perspective?

@u-fischer suggested that you can have e.g. empty H1 level (\section) with then H2 (\frametitle) inside it: but that feel a bit awkward. I'd need her to provide detail!

[u-fischer]

is it truly impossible to have an “invisible” section from a tagging perspective?

It is certainly allowed to have empty structures, and so "invisible" sections. The question is if it makes sense.
There was a bit discussion about that on one the pdf-lists and the opinions were a bit divided ;-)

I was on the side that thought it makes sense:
One main purpose imho of sectioning is to allow navigation, that means to jump from section to section etc. User of a presentation wants to jump from slide to slide and from section to section, so a slide should be a sectioning unit with the frametitle the section title (possibly empty), and sections similar.

[josephwright]

OK, so that suggests that every slide should have a \frametitle even if empty (conceptually - not a required part of the input) (I see that PowerPoint includes Slide X in the PDF structure, which neatly avoids empty name). But I'm not sure it addresses the section level of \frametitle directly.

[samcarter]

Will it be a problem for tagging if users then place higher-level sectioning (e.g. a section) within the frame? This is mostly user-error, but some tools, like biblatex, will do this by default.

[josephwright]

Not currently decided I guess :) If we do resolve that \frametitle maps to some section level, then we either suppress \section, etc., or we line thing up so \section produces \frametitle. In many ways, suppressing is perhaps cleaner - would presumably go with a warning.

[Skillmon]

I don't have an answer to this (just an opinion which is to show sections in the frame title and subsections in the frame subtitle (possibly with an addition, so something like <section>: <frametitle> as the contents of the user-perceived frame title if both were given and likewise <subsection>: <framesubtitle> as the user-perceived frame subtitle if both were given)), but that's just how I created presentations in the past using a customised beamer-derived class (not that I did many presentations during my short academic life).

However I want to also highlight that any decision on using or not using \sections might also affect article or handout mode, not only presentations.

[josephwright]

We are pretty clear that whilst we probably need a way of tagging presentations, the most logical place to provide a tagged version is likely a handout form of the content - overlays, etc., otherwise mean repeated material in a way that's unlikely to be useful. So in many ways the question is 'What does \frametitle do in a handout?'.

In terms of article mode, I am pretty wary of it - I see why Till provided it, but in reality what you need for a presentation (even with handout notes) and what you need for an article are quite different. The technical side is easy (mainly absorbing <> and throwing away), it's more the concept that worries me.

---

I see the idea of combining section and frame title, but if we are looking at supporting existing content, there will be a lot with sections in where it's not expected these appear in the title of every slide. Also, is the logical section then the \section argument or both parts (in which case every frame is a different section).

[davidcarlisle]

Mostly an outsider view (as I haven't used beamer much) but I doubt you can decide this in general, but from my limited use I'd suggest it has to be an author opt-in so probably make \section ignored in presentations by default but have a top level command (that could be used by authors or set in some themes) \UseSectionsAsFrametitle or whatever

[josephwright]

You mean I guess the question of whether section names go into slide titles? That would I guess be a theme thing.

[cfr42]

Skillman asked me to comment on this. The following is highly opinionated. Bear in mind that I've seen the code for probably one presentation I didn't write (which Joseph gave me). I've also never liked this in Beamer.

I'm not going to change this, but I'm not going to paste it either.

structure.txt

[Skillmon]

@cfr42 thanks for writing that up, unfortunately you seem to be just as lost as I am :) But I guess that's every ones current state regarding this.

@josephwright can't we put the sections in the embedded ToC (or is that what "just for the bookmarks" means)? So if you view the PDF's outline (many viewers have such a feature) you see the sections and the slides (with their titles) are grouped as subentries of the current most lowlevel sectioning; making slides always be one level below the current sectioning level of traditional sectioning commands.

[cfr42]

@cfr42 thanks for writing that up, unfortunately you seem to be just as lost as I am :) But I guess that's every ones current state regarding this.

sorry about that. It is precisely the being lost which I've never liked. I've never used powerpoint, but beamer has always seemed to encourage distracting/cluttered slides and overly complicated structures. but, yes, of course I'm not less lost than you. what did you expect?! ;)

@josephwright can't we put the sections in the embedded ToC (or is that what "just for the bookmarks" means)? So if you view the PDF's outline (many viewers have such a feature) you see the sections and the slides (with their titles) are grouped as subentries of the current most lowlevel sectioning; making slides always be one level below the current sectioning level of traditional sectioning commands.

I'd favour making them part of the structure somehow, even if not every slide has one. if you look at a set of slides, it's natural (I think) to read the frame titles as headings of some kind. this is true of powerpoint slides, too. I would have thought that this needed to be reflected in the structure provided to a screen-reader because you'd want it to be able to communicate that visual effect non-visually?

how accessible are powerpoint slides? do we know that's a good model? note that I've never used it, so I really have no clue here.

[josephwright]

I'd favour making them part of the structure somehow, even if not every slide has one. if you look at a set of slides, it's natural (I think) to read the frame titles as headings of some kind. this is true of powerpoint slides, too. I would have thought that this needed to be reflected in the structure provided to a screen-reader because you'd want it to be able to communicate that visual effect non-visually?

how accessible are powerpoint slides? do we know that's a good model? note that I've never used it, so I really have no clue here.

Well more than beamer ones :) I think one issue here is that at least to me it's not clear that slides are that accessible - certainly my ones are very visual with little in the way of text. Some structure is possible, but whether it's going to be as good as a parallel set of notes, I'm less sure.

[Skillmon]

but, yes, of course I'm not less lost than you. what did you expect?! ;)

You, and @josephwright, both created and held much more presentations than I did with beamer (my count is 4 currently, but I spend much time creating a template for that because I was the first student allowed to use LaTeX for everything there and wanted others to be allowed as well, hence I created classes and templates and whatnot for that). So I was hoping that you can bring some insides to the table. But don't you worry, the fact that everyone seems as clueless as I am is somehow soothing.

I'm afraid we don't find a strong consensus on whether to display sectioning commands on the slides or not (and moving that decision into the templates/themes makes template switching non-trivial).

[josephwright]

I'm afraid we don't find a strong consensus on whether to display sectioning commands on the slides or not (and moving that decision into the templates/themes makes template switching non-trivial).

No, but I think in the end we will have to make a decision with guidance: it's clear that accessible content will need some rules. And I agree entirely about the templates point: we will need consistency here.

[josephwright]

@FrankMittelbach suggests that \frametitle is a visual element, and structure should be handled using \section, etc. The argument to \frametitle would then be treated as an artefact - handy for the case of plain slides and so on (basically means one ignores the non-main-area text). What it would do is require users to rethink there structures - but I suspect that's true anyway for accessible PDFs.

[josephwright]

OK, I've been watching a lot of presentations over the past day and also thinking a bit about this. If you look at the structure

\begin{frame}
  Stuff here
\end{frame}

everything inside the frame environment ends up directly in the 'main content' area of the output: the bit we are clear we need to tag. That's (probably) not the long-term aim for \frametitle, which feels like it lives in the header - just not implemented yet that way.
I wonder if therefore that lines up with what @FrankMittelbach already suggested: that \frametitle is a purely visual thing and we should look to \section to be the formal structural element. One might then have a situation where the \section (and possibly \subsection) is typeset in the header unless a \frametitle is also present; that would of course depend on the theme (once we get to those).

This then means that

\section{Some stuff}
\begin{frame}
  Stuff here
\end{frame}

\begin{frame}
  More stuff here
\end{frame}

has two frames with the same 'name', but that's quite commonly how my own content works - except at the moment I use \frametitle.

(Lots of people don't use frame titles at all for presentations - that would be fine as the theme would suppress output, not alter logical strucutre.)

[josephwright]

In #12 (reply in thread), @FrankMittelbach suggests we could tag frame titles as a Caption, cf. the names of theorems, _etc.: that might work quite well and does mean we have a parallel structure for sections.

[cfr42]

so if you had an actual caption, it might be a sub-sub-caption? frame title > frame subtitle > caption?

do your frame titles match your sections? I most typically use sections as frame titles and subsections as subtitles, which may either reflect or explain my tendency to think of them structurally.

[josephwright]

Oh no, captions in presentations ....

[FrankMittelbach]

What do you mean by "an actual caption" @cfr42 ? When I said tagged as Caption I meant the PDF structure, not something the user supplies such as \caption(not that this is support in frames is it?)

E.g. in \begin{theorem}[Fermat] the string "Theorem 1. (Fermat)" is tagged inside the PDF with <Caption> ...

[josephwright]

What do you mean by "an actional caption" @cfr42 ? When I said tagged as Caption I meant the PDF structure, not something the user supplies such as \caption(not that this is support in frames is it?)

You can have a figure in beamer - but they don't float. That's really there for article mode - which I'm really not keen on.

[FrankMittelbach]

ok, a bit odd but that really has nothing to do with my tagging suggestion. something like this (if it should/would be supported) would then end up as <Figure><Caption> ...</Caption> ... <Figure> inside whatever tag structure a frame is.

[u-fischer]

I don't have a strong opinion regarding Caption or some Hn for \frametitle (my hunch would be that Hn give better a navigation behaviour, but that can be tested). Apart from this imho something like this here could be a good start. It duplicates lots of the content currently, but everything is there, and one can experiment with options to hide content or jump to relevant slides later.

\DocumentMetadata{lang=en-GB,testphase=latest}
\documentclass{xbeamer}
\usepackage{hyperref}


\makeatletter\ExplSyntaxOn
\cs_set_protected:Npn \__xbeamer_slide_begin:
  {
    \int_gincr:N \g__xbeamer_slide_int
    \int_gset:Nn \g__xbeamer_decode_int { 1 }
    \bool_gset_false:N \g__xbeamer_slide_continue_bool
    \exp_args:Ne\tagstructbegin{tag=slide,title=Slide~\int_use:N\g__xbeamer_slide_int}             
    \vbox_set:Nw \l__xbeamer_slide_box
  }
\cs_set_protected:Npn \__xbeamer_slide_end:
  {
    \vbox_set_end:
    \vbox_set_to_ht:Nnn \l__xbeamer_slide_box { \textheight }
      {
        \use:c { __xbeamer_slide_align_ \l__xbeamer_frame_alignment_tl :n }
          { \vbox_unpack_drop:N \l__xbeamer_slide_box }
      }
    \box_use_drop:N \l__xbeamer_slide_box
    \tagstructend  
  }

\RenewDocumentCommand \frametitle { D <> { all } O {#3} m }
  {
   \par
   \tagstructbegin{tag=frametitle,title= #3}%
    \tagmcbegin{}
    % this should also use the frame number!
    \MakeLinkTarget*{slide.\int_use:N\g__xbeamer_slide_int}
    \bookmark[dest=\@currentHref,rellevel=1,keeplevel]%
      {#3~(Slide~\int_use:N\g__xbeamer_slide_int)}%
    {\tagpdfparaOff\Large\bfseries #3\par\bigskip} %format 
    \tagmcend
    \tagstructend
  }
 \ExplSyntaxOff 
\tagpdfsetup{
  role/new-tag=frame/Sect,
  role/new-tag=frametitle/Caption, %or H1 or Title?
  role/new-tag=slide/Div % or perhaps Sect?
  } 
   
\AddToHook{env/frame/begin}{\par\tagstructbegin{tag=frame}}
\AddToHook{env/frame/end}  {\par\tagstructend}
\begin{document}

%fake \section{Introduction}
\tagstructbegin{tag=Sect,title=Introduction} 
% fake next frame destination for now 
\bookmark[dest=slide.1]{Introduction}

\begin{frame}
  \frametitle{Some things I did}
  \begin{itemize}[overlay-spec = +-]
    \item Get cape
    \item Wear cape
    \item Fly!
  \end{itemize}
\end{frame}

%end of section:
\tagstructend
\end{document}

[josephwright]

I've adjusted the frame setup to use this - I'm not sure how to do the section tagging (could do with a clue on how best to add the \tagstructend as part of just \section, etc.)

[josephwright]

Assuming we decide that \frametitle belongs in the header area, I need to work out how to keep the tagging correct.

[u-fischer]

Well as we have no \begin{section} ... \end{section}, you basically have to (similar to item bodies) close a Sect when the next section starts, and at end document close the last one. That is more or less what the latex-lab-sec code is doing. Here this is imho rather straightforward as the \section doesn't directly print something, is outside the frame units, and you have only a few levels.

Regarding the frametitle structure: do not forget that you can move structures around and attach them to every parent you want. You can use something like \tagstructbegin{tag=frametitle,title= #3,parent=somestructure}, you can also ensure that it is e.g. the first structure in its parent (that is what we are doing for \caption). It doesn't really matter where people issue the \frametitle command and if it is printed in the header, footer or the center of the frame, you can always store it as first child structure of the frame structure.