Method and system for verification by reading

Patent metadata

Field	Value
Publication number	US-9767388-B2
Application number	US-201414508382-A
Country	US
Kind code	B2
Filing date	Oct 7, 2014
Priority date	Mar 26, 2014
Publication date	Sep 19, 2017
Grant date	Sep 19, 2017

Abstract

An improved method for verifying whether a character-recognition technology has correctly identified which characters are represented by character images involves displaying the uncertain character images in place of their respective hypothesis characters in a document being read a verifier. The verifier may mark incorrectly spelled words containing the uncertain character images. Based on the markings, a system adjusts a confidence level associated with the hypothesis about the uncertain character in order to obtain a confirmed hypothesis linked to the uncertain character.

First claim

What is claimed is:

1. A method comprising:
receiving, at a processor, a set of uncertain characters obtained as a result of a recognition process of a text image, the set of uncertain characters including an image of an uncertain character, a hypothesis about the uncertain character, and a confidence level associated with the hypothesis;
causing, by the processor, a display device to present to a user the image of the uncertain character from the set of uncertain characters over a text readout;
receiving, at the processor from a user input, marking data for the uncertain character; and
adjusting, using the received marking data, the confidence level associated with the hypothesis about the uncertain character for obtaining a confirmed hypothesis linked to the uncertain character.

2. The method of claim 1 , wherein the marking data is one of marked, unmarked or rejected the method further comprising:
in response to determining that the marking data is unmarked for the uncertain character, the processor increasing the confidence level of the hypothesis about the uncertain character;
in response to determining that the marking data is marked for the uncertain character, the processor decreasing the confidence level of the hypothesis about the uncertain character; and
in response to determining that the marking data is rejected for the uncertain character, the processor increasing the confidence level of the hypothesis about the uncertain character.

3. The method of claim 1 , wherein the marking data is one of marked, unmarked, or rejected, the method further comprising:
in response to determining that the marking data is unmarked for the uncertain character, the processor increasing the confidence level of the hypothesis about the uncertain character by a first amount; and
in response to determining that the marking data is rejected for the uncertain character, the processor increasing the confidence level of the hypothesis about the uncertain character by a second amount, wherein the second amount is larger than the first amount.

4. The method of claim 1 , further comprising, determining, prior to causing the image of the uncertain character to be presented, whether the confidence level associated with the hypothesis about the uncertain character is below a predefined non-zero threshold value, wherein the image of the uncertain character is caused to be presented in response to determining that the confidence level of the hypothesis about the uncertain character is below the threshold value.

5. The method of claim 4 , further comprising:
receiving an indication of a desired level of accuracy for recognition of text data; and
adjusting the threshold value in accordance with an indicated desired level of accuracy.

6. The method of claim 1 , wherein presenting the image of the uncertain character over the text readout comprises identifying the hypothesis character for the uncertain character in the text readout to enable inserting the image of the uncertain character over the identified hypothesis character in the text presented thereto.

7. The method of claim 1 , wherein the text image is different than the text readout over which the image of the uncertain character is presented thereto.

8. The method of claim 1 , wherein the marking data received from a device indicates a number of times that the image of the uncertain character was selected at the device, the method further comprising:
in response to receiving marking data indicating that the image was not selected, the processor designating the marking data for the uncertain character as unmarked;
in response to receiving marking data indicating that the image was selected an odd number of times, the processor designating the marking data for the uncertain character as marked; and
in response to receiving marking data indicating that the image was selected a non-zero even number of times, designating the marking data for the uncertain character as rejected.

9. The method of claim 1 , wherein the image of the uncertain character is inserted in place of a hypothesis character image within a word in the text readout, and wherein the marking data is indicative of whether the word as a whole was marked.

10. The method of claim 1 , wherein the presented image of uncertain character replaces a hypothesis character image within the word.

11. The method of claim 1 , wherein the processor is part of a verification device, facilitating transmission of the adjusted confidence value to a remote server system.

12. The method of claim 1 , further comprising determining whether the confidence level associated with the hypothesis about the uncertain character is resolved as true, resolved as false, or not resolved;
in response to determining that the confidence level associated with the hypothesis about the uncertain character is not resolved, repeating the steps of (a) causing a display device to present the image of the uncertain character over a text readout, (b) receiving marking data for the uncertain character, and (c) adjusting the confidence level associated with the hypothesis about the uncertain character;
in response to determining that the confidence level associated with the hypothesis about the uncertain character is resolved as true, storing verified recognized text; and
in response to determining that the confidence level associated with the hypothesis about the uncertain character is resolved as false, moving to check a next hypothesis about the uncertain character.

13. The method of claim 1 , wherein the processor is part of a display device that is remotely connected to a server system.

14. The method of claim 1 , further comprising: determining font characteristics associated with the image of the uncertain character; and adjusting font characteristics of the text readout in accordance with the determined font characteristics associated with the image of the uncertain character.

15. The method of claim 1 , further comprising determining font characteristics associated with the text readout, wherein the uncertain character is chosen from a set of uncertain characters in accordance with the uncertain character being determined most similar to the determined font characteristics associated with the text readout.

16. A non-transitory computer-readable medium having stored thereon instructions executable by a processor to cause the processor to perform functions, the functions comprising:
receiving a set of uncertain characters obtained as a result of a recognition process of a text image, the set of uncertain characters including an image of an uncertain character, a hypothesis about the uncertain character, and a confidence level associated with the hypothesis;
causing a display device to present to a user the image of the uncertain character from the set of uncertain characters over a text readout;
receiving from a user input marking data for the uncertain character; and
adjusting, using the received marking data, the confidence level associated with the hypothesis about the uncertain character for obtaining a confirmed hypothesis linked to the uncertain character.

17. The computer-readable medium of claim 16 , wherein the marking data is one of marked, unmarked or rejected the functions further comprising:
in response to determining that the marking data is unmarked for the uncertain character, increasing the confidence level of the hypothesis about the uncertain character; in response to determining that the marking data is marked for the uncertain character, decreasing the confidence level of the hypot

Assignees

• Abbyy Dev Llc

Inventors

• Pakhchanian Aram Bengurovich
• Pogosskiy Michael Pavlovich

Classifications

• G06V30/1456 Primary

  based on user interactions · CPC title

• G06V30/19167

  Active pattern learning · CPC title

• G06F18/2178

  based on feedback of a supervisor · CPC title

• G06V30/10

  Character recognition · CPC title

• G06K9/2081

  Physics · mapped topic