Export syntax

[gcelano]

I would like to export syntactic information via CSVExporter [1]. In the "Annotation Keys" field, I specify "dep" or "dep_fnc", which should identify syntax in my data, but they do not work. What am I missing here?

---

[1] https://annis.varro.informatik.uni-leipzig.de/?id=c0d8fac2-4ede-4333-84e7-9209c09b03dd#_q=bGVtbWE9InByYWVzYWdpdW0i&ql=aql&_c=b2xhX3YwLjIuMF8yLG9sYV92MC4yLjBfMQ&cl=1&cr=1&s=0&l=10&

[thomaskrause]

The current CSV exporter does not support exporting edge annotations, which is a limitation of how the internal implementation works and can't easily be changed.

[gcelano]

Is this a limitation for the CSV exporter only or for all of them?

[thomaskrause]

Unfortunately, none of the current exporters implements it. Most of them are based on the BaseMatrixExporter which can handle table of matched nodes but not the relations between them.

After a restructuring after the graphANNIS backend was added, the exporter have access to the SaltXML graph. So one could write a new exporter that knows how to identify what edges are relevant for the export. But this can be very corpus specific and hard to make generic.

Thus, we used the Python or Webservice API internally when we needed it. We wrote a script that first get the list of matches (e.g. via the /search/find endpoint, and then get the subgraph for each of these matches (e.g. with the /corpora/{corpus}/subgraph endpoint). Then the script could extract the needed information from the subgraph.

The Python-API tutorial contains an example for how to get the subgraph: https://korpling.github.io/graphANNIS/docs/v3/embed-python.html