pos support in fcs

[Eduard Drenth]

Not sure if this is about clarin-d

in fcs 2.0 draft pos tag support relies on ​http://universaldependencies.org/docs/u/pos/index.html.

Does that include ​http://universaldependencies.org/docs/u/feat/index.html?

[Dirk Goldhahn]

For FCS 2.0 we agreed on basic support only for the POS itself.

But maybe Oliver has more information on this!?

[Eduard Drenth]

If this is true we miss amongst others gender, number, person and tense. I think it would be good to support feat

[Dirk Goldhahn]

However, the FCS is only supposed to be an entry point and to give a first impression of the data available. The transformation process from internal format to the universal representation for all annotated data can be quite cumbersome for the institutions providing the data.
If queries get more specific, using the more sophisticated query system of the data providing institution itself is recommended.
But thank you for letting us know of the demand for querying such features.

[Oliver Schonefeld]

There was a long discussion about POS support in FCS and people agreed on using the Universal Dependencies (UD-17) tag set without Universal features. The choice is to be understood as a more or less (lowest) common denominator between the heterogeneous resources in CLARIN centers.

It's the business of the Endpoint provider to create a mapping from UD-17 to their local POS tagset (analyzing the FCS-QL query) and from the local tagset UD-17 (for the results). Of course, centers could also opt to re-tag their data with UD-17. There is a software component contributed (FCSannotran) by Peter Beinema from MPI Nijmegen to help with the mapping process, however the mapping tables must be created by the center (if none exist, that can be reused).

If Universal Features become part of FCS is to be decided in the future, e.g. by the Tast-Force or CLARIN-ERIC.