Post-filtering of named entities with machine learning

1. A computer-implemented method, comprising:

recognizing a candidate named entity within a text;

extracting a first subset of the text that contains the candidate named entity;

analyzing, with a classifier, one or more features of the first subset of the text to identify an error associated with the candidate named entity; and

correcting the error associated with the candidate named entity.

2. The method of claim 1 , further comprising:

storing a document image in a memory; and

recognizing the text from the document image.

3. The method of claim 1 , wherein the error associated with the candidate named entity is that the candidate named entity is not a named entity and correcting the error associated with the candidate named entity includes removing the candidate named entity as a potential named entity in the text.

4. The method of claim 1 , wherein the classifier analyzes the one or more features using a first machine learning model.

5. The method of claim 4 , wherein the first machine learning model includes one or more of a recurrent neural network, a convolutional neural network, a conditional random field model, and a Markov model.

6. The method of claim 4 , further comprising:

receiving a labeled training text comprising (i) a candidate training named entity, (ii) a subset of the training text associated with the candidate training named entity, and (iii) a labeling output indicating whether the candidate training named entity is a named entity;

creating a training feature vector, wherein the training feature vector includes one or more features of the training subset of the subset of the training text;

analyzing the training feature vector using the first machine learning model to generate an indication of whether the first machine learning model identified an error associated with the candidate training named entity;

comparing the indication with the labeling output to identify one or more errors in the indication; and

updating one or more parameters of the first machine learning model based on the one or more errors in the indication.

7. The method of claim 6 , wherein the first machine learning model is initially configured to identify errors associated with candidate named entities recognized from a first document type and updating one or more parameters of the first machine learning model enables the first machine learning model to identify errors associated with candidate named entities recognized from a second document type.

8. The method of claim 1 , wherein the candidate named entity is recognized using a second machine learning model.

9. The method of claim 1 , wherein the one or more features include at least one of a named entity label associated with the candidate named entity, a recognition accuracy prediction of the candidate named entity, a distance measure between the first subset of the text and a previous subset of the text, a distance measure between the first subset of the text and a subsequent subset of the text, an embedding vector associated with the first subset of the text, semantics of the first subset of the text, and/or a similarity of the candidate named entity contained within the first subset of the text and a named entity and/or a candidate named entity contained within a second subset of the text.

10. The method of claim 1 , wherein removing the candidate named entity improves the accuracy of named entities recognized within the text.

11. The method of claim 1 , wherein the steps of the method are performed on a plurality of candidate named entities recognized within the text.

12. The method of claim 1 , wherein the recognizing a candidate named entity within the text further comprises recognizing a plurality of candidate named entities from a plurality of texts, and wherein performing the method corrects errors associated with at least a subset of the plurality of candidate named entities from the plurality of texts.

13. A system comprising:

a processor; and

a memory storing instructions which, when executed by the processor, cause the processor to:

recognize a candidate named entity within a text;

extract a first subset of the text that contains the candidate named entity;

analyze, with a classifier, one or more features of the first subset of the text to identify an error associated with the candidate named entity; and

correct the error associated with the candidate named entity.

14. The system of claim 13 , wherein the instructions further cause the processor to:

storing a document image in a memory; and

recognizing the text from the document image.

15. The system of claim 13 , wherein the error associated with the candidate named entity is that the candidate named entity is not a named entity and correcting the error associated with the candidate named entity includes removing the candidate named entity as a potential named entity in the text.

16. The system of claim 13 , wherein the classifier analyzes the one or more features using a first machine learning model.

17. The system of claim 16 , wherein the instructions further cause the processor to:

receive a labeled training text comprising (i) a candidate training named entity, (ii) a subset of the training text associated with the candidate training named entity, and (iii) a labeling output indicating whether the candidate training named entity is a named entity;

create a training feature vector, wherein the training feature vector includes one or more features of the training subset of the subset of the training text;

analyze the training feature vector using the first machine learning model to generate an indication of whether the first machine learning model identified an error associated with the candidate training named entity;

compare the indication with the labeling output to identify one or more errors in the indication; and

update one or more parameters of the first machine learning model based on the one or more errors in the indication.

18. The system of claim 17 , wherein the first machine learning model is initially configured to identify errors associated with candidate named entities recognized from a first document type and updating one or more parameters of the first machine learning model enables the first machine learning model to identify errors associated with candidate named entities recognized from a second document type.

19. The system of claim 13 , wherein the one or more features include at least one of a named entity label associated with the candidate named entity, a recognition accuracy prediction of the candidate named entity, a distance measure between the first subset of the text and a previous subset of the text, a distance measure between the first subset of the text and a subsequent subset of the text, an embedding vector associated with the first subset of the text, semantics of the first subset of the text, and/or a similarity of the candidate named entity contained within the first subset of the text and a named entity and/or a candidate named entity contained within a second subset of the text.

20. A non-transitory, computer readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to:

recognize a candidate named entity within a text;

extract a first subset of the text that contains the candidate named entity;

analyze, with a classifier, one or more features of the first subset of the text to identify an error associated with the candidate named entity; and

correct the error associated with the candidate named entity.