It is currently not possible to dedup MARC and Dublin Core records if the title field contains non-filing characters.

Primo VE can perform deduplication between MARC records in Alma and DC records that are ingested via discovery import profiles. However, the dedup title keys that are generated for MARC records take into account if the title has non-filing characters (such as initial articles), as indicated in the 245 second indicator. It is not possible to indicate non-filing characters for DC records. Therefore, records that would otherwise be matches for deduplication have different titles.

For example, a record with the title “The Relationship Between Humans and Dogs” would have the title key “Relationship Between Humans and Dogs” for the MARC record and “The Relationship Between Humans and Dogs” for the DC record.

My preferred fix for this would be for the title key generation for DC records to automatically remove common articles from the beginning of titles. I could also see other possible solutions, like enabling a way to remove non-filing characters from DC fields via normalization rules or having a variation of the title key for MARC records that includes non-filing characters.