Entity matching with machine learning fuzzy logic
AssigneePalantir Technologies Inc.
InventorsElliot Hirsch, Johannes Beil, Lauren Brown, Nicolas Prettejohn, Paul Baseotto
Life of this patent
Estimating the term…
Abstract
A fuzzy matching system matching data records in one or more data sets based on user-customized selection of multiple fuzzy matching algorithms. Possible matches may be displayed to a user, who provides feedback on the accuracy of the matches, which may then be used by a machine learning algorithm to update weightings and parameters of the multiple fuzzy matching algorithms, such as based on machine learning analysis of the matching results and the user feedback.
Claims20
A computerized method, performed by a computing system having one or more hardware computer processors and one or more non-transitory computer readable storage device storing software instructions executable by the computing system to perform the computerized method comprising:
- identifying, for individual properties associated with a first and second set of data records, a matching algorithm selected from a plurality of matching algorithms that are configured to output a match score indicative of likelihood of a match between property values of two data records, wherein at least some of the matching algorithms are token matching, substring searches, trigram, edit-distance, metaphone, term frequency model, initialization weighting, or phrase matching;
- for respective pairs of data records:
- determining the matching algorithms associated with respective properties of the first and second sets of data records;
- executing the plurality of matching algorithms on respective property values of the pair of data records to generate a corresponding plurality of match scores;
- and
- determining an overall match score for the pair of data records based on at least some of the plurality of match scores;
- and
- displaying a results user interface indicating at least a first candidate pair of data records having a highest overall match score.
A computing system comprising:
- a hardware computer processor;
- a non-transitory computer readable medium having software instructions stored thereon, the software instructions executable by the hardware computer processor to cause the computing system to perform operations comprising:
- identifying, for individual properties associated with a first and second set of data records, a matching algorithm selected from a plurality of matching algorithms that are configured to output a match score indicative of likelihood of a match between property values of two data records, wherein at least some of the matching algorithms are token matching, substring searches, trigram, edit-distance, metaphone, term frequency model, initialization weighting, or phrase matching;
- for respective pairs of data records:
- determining the matching algorithms associated with respective properties of the first and second sets of data records;
- executing the plurality of matching algorithms on respective property values of the pair of data records to generate a corresponding plurality of match scores;
- and
- determining an overall match score for the pair of data records based on at least some of the plurality of match scores;
- and
- displaying a results user interface indicating at least a first candidate pair of data records having a highest overall match score.
A non-transitory computer readable medium having software instructions stored thereon, the software instructions executable by a hardware computer processor to cause a computing system to perform operations comprising:
- identifying, for individual properties associated with a first and second set of data records, a matching algorithm selected from a plurality of matching algorithms that are each configured to output a match score indicative of likelihood of a match between property values of two data records, wherein at least some of the matching algorithms are token matching, substring searches, trigram, edit-distance, metaphone, term frequency model, initialization weighting, or phrase matching;
- for respective pairs of data records:
- determining the matching algorithms associated with respective properties of the first and second sets of data records;
- executing the plurality of matching algorithms on respective property values of the pair of data records to generate a corresponding plurality of match scores;
- and
- determining an overall match score for the pair of data records based on at least some of the plurality of match scores;
- and
- displaying a results user interface indicating at least a first candidate pair of data records having a highest overall match score.
Dependencies are inferred from the claim text.