Running the Embedded Metadata.js translator on
https://www.nature.com/articles/ncomms14906
returns

 "abstractNote": "Advanced LIGO has detected gravitational waves from two binary black hole mergers, plus a merger candidate. Here the authors use the COMPAS code to show that all three events can be explained by a single evolutionary channel via a common envelope phase, and characterize the progenitor metallicity and masses."

which is not the Abstract, but rather the content of og:description.

From here,

From the debug in Embedded Metadata.js, it looks like the RDF is being constructed with the correct abstract as:

http://purl.org/dc/elements/1.1/description=>During its first four months of taking data, Advanced LIGO has detected gravitational waves from two binary black hole mergers, GW150914 and GW151226, along with the statistically less significant binary black hole merger candidate LVT151012. Here we use the rapid binary population synthesis code COMPAS to show that all three events can be explained by a single evolutionary channel—classical isolated binary evolution via mass transfer including a common envelope phase. We show all three events could have formed in low-metallicity environments (Z=0.001) from progenitor binaries with typical total masses ≳160M⊙, ≳60M⊙ and ≳90M⊙, for GW150914, GW151226 and LVT151012, respectively. Advanced LIGO has detected gravitational waves from two binary black hole mergers, plus a merger candidate. Here the authors use the COMPAS code to show that all three events can be explained by a single evolutionary channel via a common envelope phase, and characterize the progenitor metallicity and masses.

So, is this a "bug" that should be fixed in RDF.js or Embedded Metadata.js or in Nature Publishing Group.js?