# US 10,769,430 B2 · Method and system for correcting fabrication in a document > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 10,769,430 B2 - Document ID: US10769430B2 - Status: Granted - Kind: B2 - Application number: 16/049,235 - Filed: 2018-07-30 - Published: 2019-12-12 - Granted: 2020-09-08 - Earliest claimed: 2018-06-12 - Assignees: Wipro Limited - Applicants: Wipro Limited - Inventors: Aayush Kumar; Manmohan Kumar; Janki Rao Upada - Web page: https://patentgrep.com/patent/US10769430B2 ## CPC - G06F 40/103: Formatting, i.e. changing of presentation of documents (automatic justification G06F40/189; automatic line break hyphenation G06F40/191) - G06V 30/418: Document matching, e.g. of document images - G06F 21/44: Program or device authentication - G06F 21/64: Protecting data integrity, e.g. using checksums, certificates or signatures - G06F 40/166: Editing, e.g. inserting or deleting - G06F 40/186: Templates - G06F 40/216: using statistical methods - G06V 10/82: using neural networks - G06V 30/412: Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables - G06V 30/413: Classification of content, e.g. text, photographs or tables - G06V 30/414: Extracting the geometrical structure, e.g. layout tree; Block segmentation, e.g. bounding boxes for graphics or text - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING ## Abstract A method and fabrication correction system for correcting fabrication in a document receives an input document from a user and identifies a reference document based on one of a plurality of categories of the input document from reference document database. The fabrication correction system detects fabrication in input document based on a predefined machine learning technique, by comparing one or more parameters associated with the input document with corresponding one or more parameters of the reference document and determines fabricated regions in the input document and a type of the fabrication in each of the one or more fabricated regions, based on one or more predefined techniques. Thereafter, one or more actions is performed in each of the one or more fabricated regions based on the type of the fabrication and the reference document, upon receiving a user input, for correcting the one or more fabricated regions. ## Claims ### Claim 1 A method of correcting a fabrication in a document, the method comprising: receiving, by a fabrication correction system, an input document from a user; identifying, by the fabrication correction system, a reference document based on a category of the input document, from a reference document database; detecting, by the fabrication correction system, fabrication indicating distortion associated with at least one of text and one or more images of the input document based on a predefined machine learning technique, by comparing one or more parameters associated with the input document with corresponding one or more parameters of the reference document; determining, by the fabrication correction system, one or more fabricated regions in the input document and a type of the fabrication in each of the one or more fabricated regions, based on one or more predefined techniques, by: identifying one or more boundaries for multiple segmented regions in the input document using Mask Region Convolutional Neural Network (R-CNN) technique; generating region-proposals indicating location and boundary of fabricated regions in the input document, by applying Region Proposal Network (RPN) technique in combination with Fast R-CNN technique on each segment in the input document; and determining the type of the fabrication in each of the one or more fabricated region based on Discrete Wavelet Transform (DWT) technique, wherein the type of the fabrication comprises variation in alignment, font, wavelength, syntax pixels, ink characteristics and opacity and nature of distortion associated with the at least one of text and one or more images of the input document; and performing, by the fabrication correction system, one or more actions in each of the one or more fabricated regions based on the type of the fabrication and the reference document, upon receiving a user input, for correcting the one or more fabricated regions. ### Claim 2 (depends on 1) The method as claimed in claim 1, wherein the category of the input document is identified based on the one or more parameters associated with the input document. ### Claim 3 (depends on 1) The method as claimed in claim 1, wherein the category comprises official document, non-official document, and social media related document. ### Claim 4 (depends on 2) The method as claimed in claim 2, wherein the one or more parameters comprises text, font of the text, pixel, height, and width associated with the at least one of text and one or more images in the input document. ### Claim 5 (depends on 1) The method as claimed in claim 1 further comprising: providing a notification to the user upon detecting the fabrication; and receiving the user input for correction of the one or more fabricated regions. ### Claim 6 (depends on 1) The method as claimed in claim 1, wherein performing the one or more actions in the one or more fabricated regions comprises generating a new input document, based on the predefined machine learning technique, by correcting each of the one or more fabricated regions using the type of the fabrication and the reference document. ### Claim 7 (depends on 1) The method as claimed in claim 1, wherein the predefined machine learning technique is Generative Adversarial Network (GAN). ### Claim 8 A fabrication correction system for correcting a fabrication in a document, comprising: a processor; and a memory communicatively coupled to the processor, wherein the memory stores processor instructions, which, on execution, causes the processor to: receive an input document from a user; identify a reference document based on a category of the input document, from a reference document database; detect fabrication indicating distortion associated with at least one of text and one or more images of the input document based on a predefined machine learning technique, by comparing one or more parameters associated with the input document with corresponding one or more parameters of the reference document; determine one or more fabricated regions in the input document and a type of the fabrication in each of the one or more fabricated regions, based on one or more predefined techniques, by: identifying one or more boundaries for multiple segmented regions in the input document using Mask Region Convolutional Neural Network (R-CNN) technique; generating region-proposals indicating location and boundary of fabricated regions in the input document, by applying Region Proposal Network (RPN) technique in combination with Fast R-CNN technique on each segment in the input document; and determining the type of the fabrication in each of the one or more fabricated region based on Discrete Wavelet Transform (DWT) technique, wherein the type of the fabrication comprises variation in alignment, font, wavelength, syntax pixels, ink characteristics and opacity and nature of distortion associated with the at least one of text and one or more images of the input document; and perform one or more actions in each of the one or more fabricated regions based on the type of the fabrication and the reference document, upon receiving a user input, for correcting the one or more fabricated regions. ### Claim 9 (depends on 8) The fabrication correction system as claimed in claim 8, wherein the processor identifies category of the input document based on the one or more parameters associated with the input document. ### Claim 10 (depends on 8) The fabrication correction system as claimed in claim 8, wherein the category comprises official document, non-official document, and social media related document. ### Claim 11 (depends on 9) The fabrication correction system as claimed in claim 9, wherein the one or more parameters comprises text, font of the text, pixel, height, and width associated with the at least one of text and one or more images in the input document. ### Claim 12 (depends on 8) The fabrication correction system as claimed in claim 8, wherein the processor: provides a notification to the user upon detecting the fabrication; and receives the user input for correction of the one or more fabricated regions. ### Claim 13 (depends on 8) The fabrication correction system as claimed in claim 8, wherein the processor performs the one or more actions in the one or more fabricated regions by generating a new input document, based on the predefined machine learning technique, by correcting each of the one or more fabricated regions using the type of the fabrication and the reference document. ### Claim 14 (depends on 8) The fabrication correction system as claimed in claim 8, wherein the predefined machine learning technique is Generative Adversarial Network (GAN). ### Claim 15 A non-transitory computer readable medium including instruction stored thereon that when processed by at least one processor cause a fabrication correction system to perform operation comprising: receiving an input document from a user; identifying a reference document based on a category of the input document, from a reference document database; detecting fabrication indicating distortion associated with at least one of text and one or more images of the input document based on a predefined machine learning technique, by comparing one or more parameters associated with the input document with corresponding one or more parameters of the reference document; determining one or more fabricated regions in the input document and a type of the fabrication in each of the one or more fabricated regions, based on one or more predefined techniques, by: identifying one or more boundaries for multiple segmented regions in the input document using Mask Region Convolutional Neural Network (R-CNN) technique; generating region-proposals indicating location and boundary of fabricated regions in the input document, by applying Region Proposal Network (RPN) technique in combination with Fast R-CNN technique on each segment in the input document; and determining the type of the fabrication in each of the one or more fabricated region based on Discrete Wavelet Transform (DWT) technique, wherein the type of the fabrication comprises variation in alignment, font, wavelength, syntax pixels, ink characteristics and opacity and nature of distortion associated with the at least one of text and one or more images of the input document; and performing one or more actions in each of the one or more fabricated regions based on the type of the fabrication and the reference document, upon receiving a user input, for correcting the one or more fabricated regions. ### Claim 16 (depends on 15) The non-transitory computer readable medium as claimed in claim 15, wherein the instruction causes the processor to identify the category of the input document based on the one or more parameters associated with the input document. ### Claim 17 (depends on 16) The non-transitory computer readable medium as claimed in claim 16, wherein the category comprises official document, non-official document, and social media related document. ### Claim 18 (depends on 16) The non-transitory computer readable medium as claimed in claim 16, wherein the one or more parameters comprises text, font of the text, pixel, height, and width associated with the at least one of text and one or more images in the input document. ### Claim 19 (depends on 15) The non-transitory computer readable medium as claimed in claim 15, wherein the instruction causes the processor to: provide a notification to the user upon detecting the fabrication; and receive the user input for correction of the one or more fabricated regions. ### Claim 20 (depends on 15) The non-transitory computer readable medium as claimed in claim 15, wherein the instruction causes the processor to perform the one or more actions in the one or more fabricated regions by generating a new input document, based on the predefined machine learning technique, by correcting each of the one or more fabricated regions using the type of the fabrication and the reference document. ### Claim 21 (depends on 15) The non-transitory computer readable medium as claimed in claim 15, wherein the predefined machine learning technique is Generative Adversarial Network (GAN).