comparators¶

class comparators.similarity.Similarity(**kwargs)¶

Provides methods to decide whether two objects are similar. All similarities that used in integration system must conform the interface of this class, and use this class as a parent.

fields¶: a list of attributes of data, to which metrics will be applied.

Similarity.are_similar(object1, object2)¶

examines if two objects are similar

Parameters

object1 – a dictionary with extracted data
object2 – a dictionary with extracted data

Returns

True if all metrics greater than the thresholds specified in the fields, otherwise false

abstract Similarity.get_similarities(object1, object2)¶

returns all similarity metrics of objects It is useful if you need to cache results of comparison

Parameters

object1 – a dictionary with extracted data
object2 – a dictionary with extracted data

Returns

dict of similarities

abstract Similarity.are_similar_based_on_scores(scores)¶

examines if scores are greater than thresholds in the fields

Parameters: scores – list of similarity scores of two objects
Returns: True if all scores greater than the thresholds specified in the fields, otherwise false

class comparators.factory.ComparatorFactory¶: Instantiates comparator according to full name of module and class passed in params

ComparatorFactory.get(params)¶

Instantiates comparator according to full name of module and class passed in params

Parameters: params – a dictionary with name - full name of module and class, and arguments to pass to that class
Returns: Instance of class that passed in name key in params, with arguments

class comparators.levenshtein_similarity.LevenshteinSimilarity(**kwargs)¶

See base class comparators.similarity.Similarity Calculates Levenshtein similarity normalized by max length of strings

fields¶: a list of attributes of data, to which metrics will be applied.

name¶: a name of attribute to which metric is applied

minimum_score¶: threshold which every metric should pass to assume that object are similar

LevenshteinSimilarity.are_similar(object1, object2)¶

examines if two objects are similar

Parameters

object1 – a dictionary with extracted data
object2 – a dictionary with extracted data

Returns

True if all metrics greater than the thresholds specified in the fields, otherwise false

LevenshteinSimilarity.get_similarities(object1, object2)¶: See base class.

LevenshteinSimilarity.are_similar_based_on_scores(scores)¶: See base class.

class comparators.agreement_disagreement.AgreementDisagreement(**kwargs)¶

See base class comparators.similarity.Similarity

fields¶: a list of attributes of data, to which metrics will be applied.

name¶: a name of attribute to which metric is applied

minimum_score¶: threshold which every metric should pass to assume that object are similar

weight¶: number which denotes the weight of attribute

temporal_coefficient¶

dictionary that contains:

span - a integer number of time span (3 months for example)
type - a string “agreement” or “disagreement”, which denotes type of metric
span_type - a string value of span type (for example: “year” or “month” or “day”)
value - the probability of event to happen

Example:

fields = [
    {
        "name": "name",
        "weight": 0.4,
        "temporal_coefficient": {
            'span': 4,
            'span_type': 'month',
            'type': 'agreement',
            'value': 0.001
        }
    },
    {
        "name": "profession",
        "weight": 0.2,
        "temporal_coefficient": {
            'span': 4,
            'span_type': 'month',
            'type': 'disagreement',
            'value': 0.8
        }
    },
    {
        "name": "home_airport",
        "weight": 0.2,
        "temporal_coefficient": {
            'span': 4,
            'span_type': 'month',
            'type': 'agreement',
            'value': 0.01
        }
    },
    {
        "name": "co_travellers",
        "weight": 0.2
    }
]
parameters = {
    'fields': fields,
    'minimum_score': "0.85",
}

AgreementDisagreement.are_similar(object1, object2)¶

examines if two objects are similar

Parameters

object1 – a dictionary with extracted data
object2 – a dictionary with extracted data

Returns

True if all metrics greater than the thresholds specified in the fields, otherwise false

AgreementDisagreement.get_similarities(object1, object2)¶

See base class.

Returns 1 score as a list

AgreementDisagreement.are_similar_based_on_scores(scores)¶

See base class.

Expects the list of scores that have 1 value