Preferred Label column in CSV exports mixes languages for multilingual ontologies

[jvendetti]

A user reported the following issue with the CSV representation of multilingual ontologies:

I have now discovered--or actually been notified by NCI--that the CSV download of RadLex seems to be randomly selecting either the English or the German preferred term (each identified in the "rdfs:label" field, with a language tag to differentiate), and then not including the other one in the file.

Steps to reproduce:

1. Download the CSV representation for the latest RADLEX submission (version 4.3).
2. Navigate to row 3 (Class ID: http://www.radlex.org/RID/RID21755).
3. Inspect the value of Column B (the Preferred Label column).
4. Open the same term in the BioPortal web interface with "All Languages" selected in the language selector.

Expected: Column B contains the English preferred label for the class, consistently across all rows.

Actual: Column B contains Truncus des Ramus pharyngeus des rechten Nervus glossopharyngeus — the German label for the class — even though the class has both English and German labels designated (visible in the web interface). Scanning down Column B shows that some classes are recorded in English and others in German, with no apparent pattern.

Proposed fix:

The reporter suggested standardizing on English:

Is it possible for the CSV file to uniformly use the English term? (If it can also include the German term, that would be great, but not as important.)

This aligns with previous internal discussion on CSV generation for multilingual ontologies and seems like a reasonable first pass. Including the non-English label(s) in additional column(s) could be a follow-up.