Similarity scores can be improved by looking at other non-matching identities

[epipav]

crowd.dev edition

Cloud (hosted)

Version

No response

Link

No response

Describe the problem

Sometimes similarity scores are too confident

Describe the improvement

Currently similarity score of merge suggestions are mainly based on the edit distance between matching identities. A better similarity score can be calculated by looking at the other non-matching identities of two members and checking if there are some clues (we can look at same-platform identities and how well they match for instance) that can improve the similarity score.

Additional context

No response