Add support for semantic markup in bibliographies

[frederik-elwert]

For some output formats, it is desirable to not only have a formatted bibliography, but to use semantic markup to identify parts of the bibliography (e.g., title, author, publisher, …). While pandoc supports this for JATS, it is lacking for other output formats like TEI (https://www.tei-c.org/release/doc/tei-p5-doc/en/html/ref-bibl.html) or HTML (https://schema.org/CreativeWork).

The old pandoc-citeproc processor supported this via raw content elements as an extension to CSL, which allowed citation styles to specify bits of semantic markup to be added to the output. An example for TEI using the old processor can be found here: https://github.com/frederik-elwert/teicite.

It would be desirable if either citeproc re-implements the CSL extension, or if it (or pandoc) provides an alternative solution for adding semantic markup to bibliographies.

[tarleb]

Could it make sense to somehow preserve the names of macros by wrapping the contents in a span with the macro's name? That would make it easy to post-process the output with a filter.

[jgm]

Can't do this with the current API. At the least, we'd need new methods for CiteprocOutput class, which would create a labeled Span in pandoc output.

[frederik-elwert]

I’m not sure preserving the macro names would be the best thing to do, as the macro names are basically up to the style authors? I guess ideally the variable names themselves would be preserved for the different output elements like names, text etc.

[YvesMarcoux]

IMHO, the best way to get a semantically rich bibliography in any kind of output is to completely bypass the application of a CSL style (or any styling mechanism, for that matter). CSL is designed (and works well) for controlling the presentation of a bibliography. Most CSL styles leave out on purpose some of the semantically-relevant information available in the bibliographic record, for the sake of readability or conformance to a bibliographic style (the most common case in point being the replacement of author names by "et al."). I am not sure one could write a CSL style that does not leave out anything. For example, I don’t think the dropping or non-dropping status of a name particle could directly and explicitly be represented in the output of a CSL style, even when that information is explicit in the input to the style (e.g. when that input is expressed in CSL-JSON).

Thus, it seems to me that the natural approach for producing semantically rich bibliographies in TEI or in Schema.org HTML (https://schema.org/CreativeWork) with pandoc is to develop specific writers.

One problem, in the case of a TEI output, is that even using the more structurally rigid biblStruct (as opposed to the less strict bibl), the number of valid structural representations for the same bibliographic information is very large. The writer would need to pick one specific representation, which might make it useful only in a limited range of use cases (i.e. TEI projects). However, different predefined structures could be supported (and selected by an option), covering more use cases (but not all).

Note that I am talking only about bibliographies. CSL processing might still be useful for handling citations and cites (in their CSL meaning).