System and method for matching and assembling records
A system and method for matching and assembling records is provided. One embodiment of the invention assembles records by applying a method for grouping records based on matching fields, assembling a new record as a composite of the matched records, and then repeating the grouping, matching and assembly steps in a cascade where the matching, grouping and assembly steps are modified as a function of the cascade step and the assembled records created in earlier steps.
1. A non-transitory computer readable medium having computer executable program code embodied thereon, the computer executable program code configured to cause a computing device to label a field within a record by:
providing an unlabeled field;
comparing the unlabeled field with a labeled field;
identifying the labeled field within the unlabeled field; and
assigning a same field label from the labeled field to the unlabeled field, wherein the field label is a concept identifier from an ontology.
2. The computer readable medium of claim 1 , wherein the ontology comprises the MesH ontology.
3. The computer readable medium of claim 1 , wherein the ontology comprises a set of concepts specified in a particular way and used to categorize and identify products and services.
4. The computer readable medium of claim 1 , wherein the ontology includes concept identifiers for medical words and phrases.
5. The computer readable medium of claim 1 , wherein the label is assigned only to a substring of the unlabeled field.
6. The computer readable medium of claim 1 , wherein comparing the unlabeled field with the labeled field comprises generating a match score that represents an amount of similarity between the unlabeled field and the labeled field.
7. The computer readable medium of claim 6 , wherein generating the match score comprises generating a numerical value for each of the unlabeled field and the labeled field.
8. The computer readable medium of claim 6 , wherein the match score is generated by a model derived via machine learning.