Unexpected nested filter behavior

General
1

I have a collection of entities that have 2 properties: ‘kind’ and ‘unique_key’.
I want to search for a specific small set of entities with these 2 properties specified for each. For example:

entities = [
    ['skill','python_programming'],
    ['job_role','biomedical_scientist'],
    ['skill','machine_learning'],
]

From this list I build a nested filter using And within OR. So each item returned must have exact match for both ‘kind’ and ‘unique_key’ The filter below works some times, but others it returns many more results without exact matches. By definition the combination of kind and unique_key is always unique in the collection. Am I making some basic mistake in this filter? This is using V3 library in python. Many thanks in advance.

{'operands': [{'operands': [{'operator': 'Equal',
                             'path': 'kind',
                             'valueText': 'skill'},
                            {'operator': 'Equal',
                             'path': 'unique_key',
                             'valueText': 'python_programming'}],
               'operator': 'And'},
              {'operands': [{'operator': 'Equal',
                             'path': 'kind',
                             'valueText': 'job_role'},
                            {'operator': 'Equal',
                             'path': 'unique_key',
                             'valueText': 'biomedical_scientist'}],
               'operator': 'And'},
              {'operands': [{'operator': 'Equal',
                             'path': 'kind',
                             'valueText': 'skill'},
                            {'operator': 'Equal',
                             'path': 'unique_key',
                             'valueText': 'machine_learning'}],
               'operator': 'And'}],
 'operator': 'Or'}
2

Hi @tolsonM !
Welcome to our community :hugs:

Sorry for the delay here :frowning:

This kind of issue is best solved when we have a python notebook where we can reproduce, as there can have other configurations that change the behavior.

For example, the tokenization of those fields would change the behavior. The default tokenization is wordand would remove non alfa-numeric characters, and directly impact on results.

Let me know if this help, before I proceed on creating a notebook.

Thanks!