frex.pipeline_stages.scorers package

Submodules

frex.pipeline_stages.scorers.candidate_bool_scorer module

class frex.pipeline_stages.scorers.candidate_bool_scorer.CandidateBoolScorer(*, success_scoring_explanation: frex.models.explanation.Explanation, failure_scoring_explanation: frex.models.explanation.Explanation, **kwargs)

Bases: frex.pipeline_stages.pipeline_stage.PipelineStage

CandidateBoolScorer is a scoring pipeline stage that scores based on whether or not the candidate matches some given condition. The score function for this class returns a bool indicating whether the condition was matched, and an appropriate explanation is attached to the candidate.

Parameters

• success_scoring_explanation – The explanation to add to the Candidate after applying the scoring function if the score function indicates True.

• failure_scoring_explanation – The explanation to add to the Candidate after applying the scoring function if the score function indicates False.

__call__(*, candidates: Generator[frex.models.candidate.Candidate, None, None], context: Any) → Generator[frex.models.candidate.Candidate, None, None]

For each of candidate being yielded by the Generator, apply a scoring function to the candidate and yield it as output. Based on whether or not the scoring function passed some condition, different explanations are applied.

Parameters

• candidates – A Generator yielding candidates. In the setup of a FREx Pipeline, this is typically another PipelineStage that is yielding candidates into the next stage.

• context – The current context being used to execute the Pipeline.

Returns

A Generator, yielding updated Candidate objects that have this stage’s scoring function applied.

abstract score(*, candidate: frex.models.candidate.Candidate) → Tuple[bool, float]

Apply a custom scoring function to the input candidate that also checks for the success of some condition.

Parameters

candidate – A domain-specific candidate to apply the scoring function

Returns

A tuple (x, y) where x is a boolean indicating whether the candidate passed some condition and y is the score applied to the candidate.

frex.pipeline_stages.scorers.candidate_ranker module

class frex.pipeline_stages.scorers.candidate_ranker.CandidateRanker

Bases: frex.pipeline_stages.pipeline_stage.PipelineStage

CandidateRanker is a PipelineStage that will sort the current candidates. Sorting is performed based on the total_score property of candidates, which sums up the applied_scores currently applied to the candidate.

Sorting needs to collect all candidates coming in from the generator, so it should be used infrequently.

__call__(*, candidates: Generator[frex.models.candidate.Candidate, None, None], context: Any) → Generator[frex.models.candidate.Candidate, None, None]

Collect all candidates being yielded by an input Generator, sort them based on their total_score, and yield the sorted candidates in descending order.

Parameters

• candidates – A Generator yielding candidates. In the setup of a FREx Pipeline, this is typically another PipelineStage that is yielding candidates into the next stage.

• context – The current context being used to execute the Pipeline.

Returns

A Generator, yielding Candidate objects in order based on their total_score property.

frex.pipeline_stages.scorers.candidate_scorer module

class frex.pipeline_stages.scorers.candidate_scorer.CandidateScorer(*, scoring_explanation: frex.models.explanation.Explanation, **kwargs)

Bases: frex.pipeline_stages.pipeline_stage.PipelineStage

CandidateScorer is a PipelineStage that applies some score to candidates. The scoring function must be defined in the particular application that is utilizing FREx.

A new CandidateScorer class can be minimally defined by creating a new subclass of CandidateScorer and defining the score() function.

Parameters

scoring_explanation – The explanation to add to the Candidate after applying the scoring function.

__call__(*, candidates: Generator[frex.models.candidate.Candidate, None, None], context: Any) → Generator[frex.models.candidate.Candidate, None, None]

For each of candidate being yielded by the Generator, apply a scoring function to the candidate and yield it as output.

Parameters

• candidates – A Generator yielding candidates. In the setup of a FREx Pipeline, this is typically another PipelineStage that is yielding candidates into the next stage.

• context – The current context being used to execute the Pipeline.

Returns

A Generator, yielding updated Candidate objects that have this stage’s scoring function applied.

abstract score(*, candidate: frex.models.candidate.Candidate) → float

Apply a custom scoring function to the input candidate.

Parameters

candidate – A domain-specific candidate to apply the scoring function

Returns

A score applied to be applied the candidate based on the implemented scoring function.

Module contents