Run an entity matching pipeline. Data is downloaded from CDF and an entity matching model is fit according to the parameters specified when the pipeline was created.

Match generation

Which matches are returned is then decided based on the sources the pipeline trusts the most:

1. If a source has a confirmed match, this match is returned.
2. If not 1, but a match is found using one of the regex-rules in rules and the match found is not in the rejected matches, this match is returned.
3. If not 1 nor 2, but the entity matching model find a match with score above the threshold, and the match found is not in the rejected matches, this match is returned.
4. If none of the above applies, no match is returned for the source.

Improving results

If additional matches are confirmed and added to the list of confirmed/rejected matches, or if additional rules are confirmed and added to rules through the update endpoint, re-running the pipeline should give new and improved results.