Cohort-based learning from user edits
AssigneeAmazon Technologies, Inc.
InventorVasant Manohar
Life of this patent
Estimating the term…
Abstract
A platform for generating a first character recognition-based work including a first plurality of automatically-made edits, each edit being characterized by a Unicode and a confidence score. The platform may identify at least one edit as being of questionable accuracy based on the confidence score, may determine a unique character signature of the edit, and may receive a manual correction made to the edit. The platform may also store the manual correction in association with the character signature and the Unicode, such that the manual correction is configured for use in generating a second plurality of automatically-made edits in a second character recognition-based work different than the first work.
Claims20
A method comprising:
- generating, by one or more computing devices, a first character recognition-based work including a first plurality of automatically-made edits made by the one or more computing devices, each edit of the first plurality of edits being characterized by a Unicode and a confidence score;
- comparing the respective confidence scores of the first plurality of automatically-made edits to a confidence score threshold;
- identifying at least one edit of the first plurality of automatically-made edits as having a respective confidence score below the confidence score threshold;
- characterizing the at least one edit of the first plurality of automatically-made edits as being of questionable accuracy based at least in part on the respective confidence score of the at least one edit being below the confidence score threshold;
- determining a character signature of the at least one edit, wherein the character signature comprises one or more of a shape identifier, a boundary identifier, or a location identifier, and wherein the character signature is indicative of a character of the at least one edit;
- receiving, from a first user of the one or more computing devices, a correction made to the at least one edit, the correction comprising one or more revised characters;
- storing, at the one or more computing devices, the one or more revised characters in association with the character signature and the Unicode of the at least one edit;
- and
- generating, using the one or more revised characters, a second plurality of automatically-made edits in a second character recognition-based work, wherein
- the second character recognition-based work is different than the first character recognition-based work.
One or more non-transitory computer-readable media storing computer-readable instructions that, when executed, instruct one or more processors to perform operations comprising:
- generating a first character recognition-based work including a first plurality of automatically-made edits, each edit of the first plurality of automatically-made edits being characterized by a Unicode and a confidence score;
- comparing the respective confidence scores of the first plurality of automatically-made edits to a confidence score threshold;
- identifying at least one edit of the first plurality of automatically-made edits as having a respective confidence score below the confidence score threshold;
- determining a character signature of the at least one edit;
- receiving a manual correction made to the at least one edit;
- storing the manual correction in association with the character signature and the Unicode of the at least one edit;
- and
- generating, using the manual correction, to generate a second plurality of automatically-made edits in a second character recognition-based work, the second character recognition-based work being different than the first character recognition-based work.
One or more computing devices, comprising:
- one or more processors;
- and
- memory, wherein the memory includes executable instructions that, when executed by the processor, cause the processor to perform acts comprising:
- generating a first character recognition-based work, the first character recognition-based work including a first plurality of automatically-made edits made by the one or more computing devices;
- assigning a Unicode and a confidence score to each edit of the first plurality of automatically-made edits;
- comparing the respective confidence scores of the first plurality of automatically-made edits to a confidence score threshold;
- identifying at least one edit of the first plurality of automatically-made edits as having a respective confidence score below the confidence score threshold, and
- associating a character signature with the at least one edit;
- receiving a manual correction made to the at least one edit;
- storing the manual correction in the memory in association with the character signature and the Unicode of the at least one edit;
- and
- generating, based at least in part on the manual correction, a second plurality of automatically-made edits in a second character recognition-based work, and wherein the second character recognition-based work is different than the first character recognition-based work.
Dependencies are inferred from the claim text.