# Embedding citation-specific metadata in PDF files

**URL:** <https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231>\
**Category:** CSL Development\
**Created:** [June 25, 2014, 5:17pm UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231 "2014-06-25T17:17:06Z")\
**Posts on this page:** 13\
**Page:** 1

<div class="post-metadata">

**Author:** ![johnmie](https://avatars.discourse-cdn.com/v4/letter/j/ea666f/32.png) [@johnmie](https://discourse.citationstyles.org/u/johnmie)\
**Post date:** [June 25, 2014, 5:17pm UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/1 "2014-06-25T17:17:06Z")

</div>

digi-libris Reader [http://digi-libris.com](http://digi-libris.com) can export all Metadata of an  
object including CSL-specific variables (those that cannot be mapped 1:1 to  
Dublin Core Terms. e.g. pageRange, event, genre etc.) as XMP sidecar file  
which can be imported into PDF files (with Acrobat.exe) and which other  
software might be able to read.

Until now we have stored these CSL variables as attribute/value pairs under  
custom entries which appear in Acrobat.exe under /File \>\> Properties \>\>  
Additional Metadata \>\> Advanced/ and are stored in the XMP file as

/\<rdf:Description rdf:about="" xmlns:pdfx=“[http://ns.adobe.com/pdfx/1.3/](http://ns.adobe.com/pdfx/1.3/)”\>  
pdfx:citation\_pageRange7-9\</pdfx:citation\_pageRange\>  
\</rdf:Description\>/

I am now considering changing this to include a proper CSL namespace which  
could look line  
/  
\<rdf:Description rdf:about="" xmlns:cs=“[http://purl.org/net/xbiblio/csl/](http://purl.org/net/xbiblio/csl/)”\>  
cs:pageRange7-9\</cs:pageRange\>  
\</rdf:Description\>/

but unfortunately this URL returns a 404 error or automatically re-directs  
you to [http://citationstyles.org](http://citationstyles.org). No way to see a list of variables.

What do you recommend:

1 stay with pdfx  
2 change to xbiblio even though the latter does not reveal a valid  
namespace  
3 register a new domain with [purl.org](http://purl.org) (under [purl.org/digi/csl/](http://purl.org/digi/csl/) or  
similar)  
4 as nbr 3 above but use a proprietary prefix (e.g. digicita: or similar)  
?–  
View this message in context: [http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100.html](http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100.html)  
Sent from the xbiblio-devel mailing list archive at [Nabble.com](http://Nabble.com).

---

<div class="post-metadata">

**Author:** ![Rintze\_Zelle](https://yyz1.discourse-cdn.com/flex029/user_avatar/discourse.citationstyles.org/rintze_zelle/32/175_2.png) [@Rintze\_Zelle](https://discourse.citationstyles.org/u/Rintze_Zelle)\
**Post date:** [July 14, 2014, 1:30am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/2 "2014-07-14T01:30:06Z")

</div>

Since nobody is responding, my two cents: I would pick option 4 for now.

Rintze

---

<div class="post-metadata">

**Author:** ![Bruce\_D\_Arcus1](https://avatars.discourse-cdn.com/v4/letter/b/4bbf92/32.png) [@Bruce\_D\_Arcus1](https://discourse.citationstyles.org/u/Bruce_D_Arcus1)\
**Post date:** [July 14, 2014, 1:47am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/3 "2014-07-14T01:47:03Z")

</div>

You could do option 5: use bibo?

[http://bibliontology.com/](http://bibliontology.com/)

---

<div class="post-metadata">

**Author:** ![Aurimas\_Vinckevicius](https://avatars.discourse-cdn.com/v4/letter/a/e47774/32.png) [@Aurimas\_Vinckevicius](https://discourse.citationstyles.org/u/Aurimas_Vinckevicius)\
**Post date:** [July 14, 2014, 1:48am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/4 "2014-07-14T01:48:12Z")

</div>

> Since nobody is responding, my two cents: I would pick option 4 for now.
> 
> Rintze
> 
> > digi-libris Reader [http://digi-libris.com](http://digi-libris.com) can export all Metadata  
> > of an  
> > object including CSL-specific variables (those that cannot be mapped 1:1  
> > to  
> > Dublin Core Terms. e.g. pageRange, event, genre etc.) as XMP sidecar file  
> > which can be imported into PDF files (with Acrobat.exe) and which other  
> > software might be able to read.

Where are these CSL-specific variables coming from? I don’t see any CSL  
spec (neither CSL documentation, nor CSL JSON format) defining pageRange.

> > Until now we have stored these CSL variables as attribute/value pairs  
> > under  
> > custom entries which appear in Acrobat.exe under /File \>\> Properties \>\>  
> > Additional Metadata \>\> Advanced/ and are stored in the XMP file as
> > 
> > /\<rdf:Description rdf:about=“” xmlns:pdfx="[http://ns.adobe.com/pdfx/1.3/](http://ns.adobe.com/pdfx/1.3/)  
> > "\>  
> > pdfx:citation\_pageRange7-9\</pdfx:citation\_pageRange\>  
> > \</rdf:Description\>/
> > 
> > I am now considering changing this to include a proper CSL namespace  
> > which  
> > could look line  
> > /  
> > \<rdf:Description rdf:about=“” xmlns:cs="[http://purl.org/net/xbiblio/csl/](http://purl.org/net/xbiblio/csl/)  
> > "\>  
> > cs:pageRange7-9\</cs:pageRange\>  
> > \</rdf:Description\>/
> > 
> > but unfortunately this URL returns a 404 error or automatically  
> > re-directs  
> > you to [http://citationstyles.org](http://citationstyles.org). No way to see a list of variables.

Namespaces are not required to resolve to a valid page (I agree that it may  
be useful though). For all intents and purposes they’re just some  
globally-unique string.

---

<div class="post-metadata">

**Author:** ![Robert\_Knight](https://avatars.discourse-cdn.com/v4/letter/r/ba9def/32.png) [@Robert\_Knight](https://discourse.citationstyles.org/u/Robert_Knight)\
**Post date:** [July 14, 2014, 5:36am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/5 "2014-07-14T05:36:08Z")

</div>

> digi-libris Reader [http://digi-libris.com](http://digi-libris.com) can export all Metadata of an  
> object including CSL-specific variables (those that cannot be mapped 1:1 to  
> Dublin Core Terms. e.g. pageRange, event, genre etc.)

Have you considered mapping to PRISM as well? [1] That fills in a  
number of gaps in Dublin Core and is already in use by several  
publishers. Mendeley will read PRISM metadata from PDFs in addition to  
Dublin Core and I think Papers does as well. I’m not sure if Zotero  
can?

[1] [http://www.prismstandard.org/specifications/2.1/PRISM\_prism\_namespace\_2.1.pdf](http://www.prismstandard.org/specifications/2.1/PRISM_prism_namespace_2.1.pdf)

---

<div class="post-metadata">

**Author:** ![Aurimas\_Vinckevicius](https://avatars.discourse-cdn.com/v4/letter/a/e47774/32.png) [@Aurimas\_Vinckevicius](https://discourse.citationstyles.org/u/Aurimas_Vinckevicius)\
**Post date:** [July 14, 2014, 6:11am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/6 "2014-07-14T06:11:22Z")

</div>

PRISM 3.0  
[http://www.idealliance.org/specifications/prism-metadata-initiative/prism/specifications/prism-30-spec](http://www.idealliance.org/specifications/prism-metadata-initiative/prism/specifications/prism-30-spec)  
has been published as well, though Zotero (I can’t speak for other  
managers) does not yet recognize the new spec/namespace, but we’ll get  
there soon. In either case, Zotero does not read metadata directly from  
PDFs, because, from what we’ve seen, the metadata is very unreliable  
(though this may change in the future).

---

<div class="post-metadata">

**Author:** ![johnmie](https://avatars.discourse-cdn.com/v4/letter/j/ea666f/32.png) [@johnmie](https://discourse.citationstyles.org/u/johnmie)\
**Post date:** [July 14, 2014, 9:21am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/7 "2014-07-14T09:21:02Z")

</div>

Thanks Rintze for relaunshing the debate

I have adopted option four as follows:  
Citation relevant variables will be stored on export in XMP sidecar files as

\<rdf:Description rdf:about="" xmlns:cs=“[http://purl.org/digilib/cita/](http://purl.org/digilib/cita/)”\>  
citation:titleEmbedded Metadata add Value to Scientific  
Publications\</citation:title\>  
citation:number-of-pages3\</citation:number-of-pages\>  
citation:original-publisher-placeGeneva\</citation:original-publisher-place\>  
…  
\</rdf:Description\>

and those which cannot be mapped to DC will also be carried under pdfx as  
custom attribute/value pairs with the ‘citation\_’ prefix, same as already in  
use in many HTML files.

to aurimas: You are absolutely right, ‘pagerange’ is not in the CSL  
specification. It is a convenience variable I have used, but it exports as  
’page’ and not as ‘pagerange’. My fault, sorry for the misleading typo.

to robert: be happy to include a bridge for PRISM variables if this is a  
widely used standard. Just show me a mapping list and the [purl.org](http://purl.org) entry to  
use.–  
View this message in context: [http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579116.html](http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579116.html)  
Sent from the xbiblio-devel mailing list archive at [Nabble.com](http://Nabble.com).

---

<div class="post-metadata">

**Author:** ![Robert\_Knight](https://avatars.discourse-cdn.com/v4/letter/r/ba9def/32.png) [@Robert\_Knight](https://discourse.citationstyles.org/u/Robert_Knight)\
**Post date:** [July 14, 2014, 9:38am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/8 "2014-07-14T09:38:18Z")

</div>

> to robert: be happy to include a bridge for PRISM variables if this is a  
> widely used standard. Just show me a mapping list and the [purl.org](http://purl.org) entry to  
> use.

I’m not sure if there is an existing [purl.org](http://purl.org) entry. The example at  
[http://www.prismstandard.org/resources/mod\_prism.html](http://www.prismstandard.org/resources/mod_prism.html) uses a  
[prismstandard.org](http://prismstandard.org) URL for the namespace. There is a PURL  
‘/rss/1.0/modules/prism/’ which points to the aforementioned  
mod\_prism.html resource but you want one which points to the  
namespace?

I don’t have a list of PRISM → CSL mappings directly to hand, but the  
fields that we recognize which I believe  
map straightforwardly to CSL in most cases are:

“prism:aggregationType”, “prism:copyright”, “prism:doi”, “prism:edition”,  
“prism:endingPage”, “prism:genre”, “prism:issn”, “prism:issueIdentifier”,  
“prism:issueName”, “prism:keyword”, “prism:location”, “prism:number”,  
“prism:organization”, “prism:pageRange”, “prism:publicationDate”,  
“prism:publicationName”, “prism:section”, “prism:startingPage”,  
“prism:volume”, “prism:url”

---

<div class="post-metadata">

**Author:** ![Robert\_Knight](https://avatars.discourse-cdn.com/v4/letter/r/ba9def/32.png) [@Robert\_Knight](https://discourse.citationstyles.org/u/Robert_Knight)\
**Post date:** [July 14, 2014, 9:29am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/9 "2014-07-14T09:29:31Z")

</div>

> In either case, Zotero does not read metadata directly from PDFs, because, from what we’ve seen,  
> the metadata is very unreliable (though this may change in the future).

The main problem we observed was that the same Dublin Core fields that  
are used for article metadata are also filled in by PDF generation  
software using generic defaults - for example the filename of the  
source document (Word, LaTeX etc.) as dc:title and the name of the  
software that created the PDF as dc:creator.

In Mendeley we apply some simple heuristics based on a comparison of  
the metadata with the actual content of the first few pages of the PDF  
to decide whether or not to use that metadata.

The presence of PRISM fields is also a useful indicator since they are  
more domain specific and less likely to be populated with other data  
than the DC fields.

---

<div class="post-metadata">

**Author:** ![Aurimas\_Vinckevicius](https://avatars.discourse-cdn.com/v4/letter/a/e47774/32.png) [@Aurimas\_Vinckevicius](https://discourse.citationstyles.org/u/Aurimas_Vinckevicius)\
**Post date:** [July 14, 2014, 5:07pm UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/10 "2014-07-14T17:07:38Z")

</div>

Probably just a typo, but your namespace declaration doesn’t match the  
prefix you are using.

More importantly, if the idea behind using this namespace URI is to offer  
interoperability between software, then I’m not sure this is helpful. I’ve  
never seen [http://purl.org/digilib/cita/](http://purl.org/digilib/cita/) namespace used in this context  
(though it may be) and I can’t find any documentation for it. Can anyone  
point to a reference?

I think your best choice for interoperability would be to use a common,  
rich vocabulary, like PRISM, with a namespace URI that is official/widely  
used (e.g. [http://prismstandard.org/namespaces/basic/2.1/](http://prismstandard.org/namespaces/basic/2.1/) or a different  
version).

If you don’t care about interoperability, then I guess it doesn’t matter at  
all.

Aurimas

> Thanks Rintze for relaunshing the debate
> 
> I have adopted option four as follows:  
> Citation relevant variables will be stored on export in XMP sidecar files  
> as
> 
> \<rdf:Description rdf:about=“” xmlns:cs=“[http://purl.org/digilib/cita/](http://purl.org/digilib/cita/)”\>  
> citation:titleEmbedded Metadata add Value to Scientific  
> Publications\</citation:title\>  
> citation:number-of-pages3\</citation:number-of-pages\>

citation:original-publisher-placeGeneva\</citation:original-publisher-place\>

> …  
> \</rdf:Description\>
> 
> and those which cannot be mapped to DC will also be carried under pdfx as  
> custom attribute/value pairs with the ‘citation\_’ prefix, same as already  
> in  
> use in many HTML files.
> 
> to aurimas: You are absolutely right, ‘pagerange’ is not in the CSL  
> specification. It is a convenience variable I have used, but it exports as  
> ‘page’ and not as ‘pagerange’. My fault, sorry for the misleading typo.
> 
> to robert: be happy to include a bridge for PRISM variables if this is a  
> widely used standard. Just show me a mapping list and the [purl.org](http://purl.org) entry  
> to  
> use.
> 
> –  
> View this message in context:  
> [http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579116.htmlOn](http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579116.htmlOn) Jul 14, 2014 4:21 AM, “johnmie” \<@johnmie\> wrote:  
> Sent from the xbiblio-devel mailing list archive at [Nabble.com](http://Nabble.com).

* * *

---

<div class="post-metadata">

**Author:** ![johnmie](https://avatars.discourse-cdn.com/v4/letter/j/ea666f/32.png) [@johnmie](https://discourse.citationstyles.org/u/johnmie)\
**Post date:** [July 14, 2014, 8:23pm UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/11 "2014-07-14T20:23:46Z")

</div>

actually it now reads \<rdf:Description rdf:about=""  
xmlns:citation=“[http://purl.org/digilib/cita/](http://purl.org/digilib/cita/)”\>, cs came from a previous  
test version.

The [purl.org/digilib/cita/](http://purl.org/digilib/cita/) link is new and the updated version of  
digi-libris reader has not yet been uploaded. This is why you have not yet  
seen it anywhere. Remember I had asked the original question only a few days  
ago and did not get any meaningful suggestions until Rintze re-launched the  
debate.

If you click on our purl link you will be directed to our Citation Variables  
appendix which documents all the variables we use and how CSL variables are  
mapped.

The prismstandard link points to an errata page which in turn returns 404.  
But from what I have seen on another prism page, many of the variables do  
not map 1:1 to either CSL or DC and will be treated as  
custom variables.

john m.–  
View this message in context: [http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579120.html](http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579120.html)  
Sent from the xbiblio-devel mailing list archive at [Nabble.com](http://Nabble.com).

---

<div class="post-metadata">

**Author:** ![Aurimas\_Vinckevicius](https://avatars.discourse-cdn.com/v4/letter/a/e47774/32.png) [@Aurimas\_Vinckevicius](https://discourse.citationstyles.org/u/Aurimas_Vinckevicius)\
**Post date:** [July 14, 2014, 8:51pm UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/12 "2014-07-14T20:51:27Z")

</div>

OK, so you’re basically establishing a new schema, which is highly parallel  
to the variables used in CSL documentation (those are effectively _not_  
under [http://purl.org/net/xbiblio/csl](http://purl.org/net/xbiblio/csl) namespace, because they are never  
used in QNAMES in that context). The only thing that is not entirely clear  
to me from the documentation is which verbs (in the RDF sense) fall under  
the “[http://purl.org/digilib/cita/](http://purl.org/digilib/cita/)” namespce. Is it all of the listed terms  
under “Variables used in CSL styles”, except for the ones marked with \* and  
\*\* (since those are mapped to dc/dcterms namespace)? Or is it all of the  
terms on that page? IMO, since you’re establishing a new namespace anyway,  
it would make sense to add all of the CSL variables to this namespace with  
no exceptions (you’re not forced to use this in digibib export anyway). I  
would also go with a namespace URI that makes this relationship clear (e.g.  
“[http://purl.org/net/xbiblio/csl-vars#](http://purl.org/net/xbiblio/csl-vars#)”, “  
[http://purl.org/net/digibib/csl-vars#](http://purl.org/net/digibib/csl-vars#)” or something similar).

---

<div class="post-metadata">

**Author:** ![johnmie](https://avatars.discourse-cdn.com/v4/letter/j/ea666f/32.png) [@johnmie](https://discourse.citationstyles.org/u/johnmie)\
**Post date:** [July 15, 2014, 8:22am UTC](https://discourse.citationstyles.org/t/embedding-citation-specific-metadata-in-pdf-files/1231/13 "2014-07-15T08:22:26Z")

</div>

digi-libris distinguishes between 5 types of variables:  
1. those that can be mapped to dc or dcterms,  
2. those that can be mapped to CSL and dc/dcterms,  
3. those that can only be mapped to CSL,  
4. those imported and re-mappable such as ris, bib, MARC21, Prism  
(planned) etc.  
5. those that cannot be mapped to any of the above (which become  
custom attribute/value pairs).

```
  In export all CSL variables are included under digilib namespace
  and in addition those also available as dc or dcterms are
  duplicated under the respective namespace, all other CSL variables are
  duplicated as custom attribute/value pairs under pdfx with the

```

‘citation\_’ prefix.  
Thus they are visible even to individuals who may not have  
adequate software to read all of the embedded metadata.–  
View this message in context: [http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579122.html](http://xbiblio-devel.2463403.n2.nabble.com/Embedding-citation-specific-metadata-in-PDF-files-tp7579100p7579122.html)  
Sent from the xbiblio-devel mailing list archive at [Nabble.com](http://Nabble.com).