Extracting values from images of documents
AssigneeLendingClub Corporation
InventorYifan Wu
Life of this patent
Estimating the term…
Abstract
Techniques are described for extracting key values from a document without having to rely on finding corresponding labels for the target keys within the extracted text of the document. Further the techniques do not rely on knowledge of the correlation between (a) the location of labels within a document, and (b) the location of the key values that correspond to the labels. Key values are extracted from a document by, identifying candidate values within the document, establishing “joint-candidate” sets from those candidate values, and using a trained machine learning mechanism to score each joint-candidate set of values. The highest scoring joint-candidate set is deemed to reflect the correct mapping of candidate values to target keys for the document.
Claims20
A method comprising:
- obtaining extracted text from a document;
- identifying a plurality of candidate values, within the extracted text, that correspond to a particular value-type;
- generating a plurality of candidate sets based on the plurality of candidate values, wherein each candidate set of the plurality of candidate sets represents a distinct mapping of candidate values, from the plurality of candidate values, to a set of target keys of the particular value-type;
- generating a respective feature set for each candidate set of the plurality of candidate sets;
- generating a plurality of scores by feeding the respective feature set of one or more candidate sets, of the plurality of candidate sets, to a trained machine learning engine to generate a respective score for each of the one or more candidate sets;
- based on the plurality of scores, selecting a particular candidate set of the plurality of candidate sets;
- and
- generating output that indicates the mapping, of candidate values to the set of target keys, that is represented by the particular candidate set;
- wherein the method is performed by one or more computing devices.
One or more non-transitory computer-readable media storing instructions which, when executed by one or more computing devices, cause:
- obtaining extracted text from a document;
- identifying a plurality of candidate values, within the extracted text, that correspond to a particular value-type;
- generating a plurality of candidate sets based on the plurality of candidate values, wherein each candidate set of the plurality of candidate sets represents a distinct mapping of candidate values, from the plurality of candidate values, to a set of target keys of the particular value-type;
- generating a respective feature set for each candidate set of the plurality of candidate sets;
- generating a plurality of scores by feeding the respective feature set of one or more candidate sets, of the plurality of candidate sets, to a trained machine learning engine to generate a respective score for each of the one or more candidate sets;
- based on the plurality of scores, selecting a particular candidate set of the plurality of candidate sets;
- and
- generating output that indicates the mapping, of candidate values to the set of target keys, that is represented by the particular candidate set.
Dependencies are inferred from the claim text.