Image reading systems, methods and storage medium for performing entity extraction, grouping and validation
AssigneeU.S. Bank National Association
InventorsSoumitri Naga Kolavennu, Raj M. Bharadwaj, Priyanka Singhal, Varshini Sriram, Lavanya Basavaraju
Life of this patent
Estimating the term…
Abstract
Entity extraction is performed using a voting ensemble approach to enhance the accuracy and effectiveness of entity extraction processes. Voting ensemble weights are stored. In turn, a prediction value for a first entity is determined from multiple models that are specifically trained for entity extraction. A plurality of first entity prediction values is obtained. The model with the highest prediction value among the plurality of first entity prediction values is then selected.
Claims20
A method for performing entity extraction, comprising:
- storing a plurality of voting ensemble weights;
- determining a first entity prediction value for each of a plurality of models trained to extract a first entity, thereby determining a plurality of first entity prediction values;
- selecting a first model having the highest prediction value of the plurality of first entity prediction values to extract a first entity:value pair type;
- extracting, from a document, an entity:value pair corresponding to the first entity using a first ensemble weight associated with the first model, the entity:value pair including the first entity:value pair type;
- and
- designating the first model to extract the first entity:value pair type from additional documents.
A system for performing entity extraction, comprising:
- a storage device storing a plurality of voting ensemble weights;
- a plurality of models trained to extract a first entity;
- one or more processors;
- and
- a non-transitory computer-readable storage medium encoding instructions which, when executed by the one or more processors, causes the one or more processors to:
- determine a first entity prediction value for each of the plurality of models, thereby determining a plurality of first entity prediction values;
- and
- select a first model having the highest prediction value of the plurality of first entity prediction values to extract a first entity:value pair type;
- and
- designate the first model to extract the first entity:value pair type from one or more documents.
A non-transitory computer-readable medium having stored thereon one or more sequences of instructions for causing one or more processors to perform:
- storing a plurality of voting ensemble weights;
- determining a first entity prediction value for each of a plurality of models trained to extract a first entity, thereby determining a plurality of first entity prediction values;
- selecting a first model having the highest prediction value of the plurality of first entity prediction values to extract a first entity:value pair type;
- and
- designating the first model to extract the first entity:value pair type from one or more documents.
Dependencies are inferred from the claim text.