Identifying versions of a form
Disclosed are a method and apparatus for identifying versions of a form. In an example, clients of a medical company fill out many forms, and many of these forms have multiple versions. The medical company operates in 10 states, and each state has a different version of a client intake form, as well as of an insurance identification form. In order to automatically extract information from a particular filled out form, it may be helpful to identify a particular form template, as well as the version of the form template, of which the filled out form is an instance. A computer system evaluates images of filled out forms, and identifies various form templates and versions of form templates based on the images.
1. A method comprising:
receiving a plurality of images of forms, each of the plurality of images of forms includes a first set of pixels corresponding to form structure that are consistent across forms of a given form version and a second set of pixels that correspond to handwriting that is unique to each of the plurality of images, each of the plurality of images is associated with one of a plurality of form versions;
dividing each of the plurality of images into sub images, each sub image is a region of a respective image and are a size that captures distinctive portions of the first set of pixels, each sub image includes metadata tying the sub image to the region of the respective image;
converting each sub image into a vector, wherein the vector describes the content of the sub image;
comparing each vector with other vectors describing a matching region of the respective image based on the metadata; and
sorting the plurality of images of forms into matching form versions having the same first set of pixels, the sorting based on said comparing each vector.
2. The method of claim 1 , further comprising:
for each category, identify the form version based on registration of a single image of the plurality of images of forms that is sorted into the respective category; and
assign all forms in each category a particular form version based on said identifying.
3. The method of claim 1 , wherein each region of the respective image included in each sub image is predetermined based on regions of the plurality of images that are most indicative of variations between the plurality of form versions.
4. The method of claim 3 , wherein the predetermined regions are based on keypoints in the plurality of images.
5. The method of claim 1 , wherein said comparison is based on a threshold of matching vector elements.
6. The method of claim 1 , wherein said comparison further includes:
assigning a class label to each vector found in the respective image.
7. The method of claim 1 , wherein said comparison is performed using a trained machine learning model that receives vectors and outputs a confidence of similarity score.
8. The method of claim 7 , wherein the confidence of similarity score is based on a “k-nearest neighbors” classification.
9. A system comprising:
an image processor configured to receiving a plurality of images of forms, each of the plurality of images of forms includes a first set of pixels corresponding to form structure that are consistent across forms of a given form version and a second set of pixels that correspond to handwriting that is unique to each of the plurality of images, wherein each of the plurality of images is associated with one of a plurality of form versions, the image processor further configured to divide each of the plurality of images into sub images, each sub image is a region of a respective image and are a size that captures distinctive portions of the first set of pixels, each sub image includes metadata tying the sub image to the region of the respective image, and the image processor is configured to convert each sub image into a vector, wherein the vector describes the content of the sub image; and
a memory including a trained machine learning model configured to compare each vector with other vectors describing a matching region of the respective image based on the metadata and sort the plurality of images of forms into matching form versions having the same first set of pixels, the sorting based on said comparing each vector.
10. The system of claim 9 , wherein the image processor is further configured to identify the form version based on registration of a single image of each category the plurality of images of forms are sorted into by the machine learning model and assign all forms in each category a particular form version based on the identification.
11. The system of claim 9 , wherein each region of the respective image included in each sub image is predetermined based on regions of the plurality of images that are most indicative of variations between the plurality of form versions.
12. The system of claim 11 , wherein the predetermined regions are based on keypoints in the plurality of images.
13. The system of claim 9 , wherein said comparison is based on a threshold of matching vector elements.
14. The system of claim 9 , wherein said comparison further includes assignment of a class label to each vector found in the respective image.