Getting top unidentified places from Recogito DB

gistfile1.sql

SELECT 
  -- concat (toponym, toponym_corrected), count(*) 
  coalesce(toponym, toponym_corrected), count(*) 
  
FROM annotations WHERE
 (status = 'NOT_IDENTIFYABLE' OR status = 'NO_SUITABLE_MATCH' OR status = 'AMBIGUOUS' OR status = 'MULTIPLE')
-- GROUP BY concat(toponym, toponym_corrected)
GROUP BY coalesce(toponym, toponym_corrected)
ORDER BY count desc ;

This query pulls the top unidentified toponyms out of our Recogito database. Toponyms are kept in a table with the following structure:

toponym     | toponym_corrected  | status
---------------------------------------------------------
Lybia       |                    | AMBIGUOUS
Africa      |                    | AMBIGUOUS
            | Metellus           | NO_SUITABLE_MATCH
Ilicitanian | Ilicitanian Gulf   | NOT_IDENTIFYABLE

The problem: it's a bit of a hack. Recogito retains toponyms (which are typically identified by an automatic process) and manual corrections to them in separate columns (so that we can later benchmark the system). For a clean query, we'd need to flatten both columns into a toponym_merged column, according to the rule

"use toponym_corrected if available, otherwise use toponym if avaliable, otherwise leave blank".

I frankly don't know how to implement this in SQL. So I'm just concatenating both columns right now (which works most of the time since the cases where both colums are populated are rare). But any SQL hints would be greatly appreciated!

As Hugh Cayless has pointed out:

In any case, the answer is probably the COALESCE() function instead of CONCAT()

Thanks again!