Dara O.

We have a dataset where we are seeing that identical records don't have identical matching min and max scores. Is this expected? For the matching algorithm, we would expect that if A == B with A, B, C, D being in the same cluster, we would expect A and B to have the same matching score with C and A and B having the same matching score to D. So A and B should have the same min and max score but we have found a few instances where they don't have the same scoring...

Good Morning! Can we dynamically change the MatchType without retraining the algorithm? Getting this error: RESOURCE_NOT_FOUND_ERROR; Failed Line Number: 61; An error occurred while calling o454.execute. z_sim26 does not exist.

Getting some awful skews when trying to scale up to 20m records. I've tried:

Is this expected? Are there any ideas for what we could do to reduce this skew?

Hello, does anyone have any ideas for the matching score and applying an automated threshold to the matches that exclude "weak" matches. We can currently apply a threshold to the zmin/zmax score but if there is a weak match within the group the min/max score is affected greatly. Is there a good way to infer what records might be strong matches to each other within a group?

Hello! Can more than 2 records have the same min/max score? Sometimes, when there are around 4 records in group they sometimes all share the same minimum and maximum score which seems strange. Is the second phase of matching a 1 to 1 or can there be one to many matches. For 1 to 1 matches I would not expect more than 2 records to share the same minimum/maximum score as each pair would have a unique score.