DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claims 1, 11, and 20, these claims recite “a plurality of text boxes associated with the text and corresponding coordinates associated with each of the text boxes”. Note that “the text” has improper antecedence and is being interpreted as “the text associated with the document”.
Claims 2-10 and 12-19 are rejected as dependent on the above claims.
Claims 6 and 16 recite “each document provider” which has improper antecedence. Further, it is unclear how a similarity of a vector is measured to non-vectors (each document provider). Therefore the claim is being interpreted such that the vector is being compared to vectors representing other document providers.
Claims 7-8 and 17-18 are rejected as dependent on the above claims.
Claims 8 and 18 recite, “wherein the document provider associated with the document is determined to be the document provider that is most similar to the vector within the document provider space.” The underlined element lacks antecedent basis and is being interpreted as referring to one document provider from a plurality of document providers from the comparison described in claims 6-7.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101.
Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea of determining a document provider, without significantly more.
The claim recites: A method, comprising:
receiving a document from a document provider;
applying optical character recognition to the document to determine text associated with the document;
determining, using a processor, a plurality of text boxes associated with the text and corresponding coordinates associated with each of the text boxes;
determining the document provider associated with the document based on the plurality of text boxes; and
updating a database to associate the document with the document provider.”
The limitations, as drafted, are processes that, under their broadest reasonable interpretation, cover performance of the limitation in the mind. A person can recognize text on a document, determine text boxes and coordinates, determine a document provider based on the text boxes, then mentally store information. The document receipt amounts to insignificant, extra-solution activity (data collection).
This judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of applying OCR and a using database. The use of OCR and a database amount to highly generic, well-known, and routinely used processes/elements. Accordingly, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. The claim is directed to an abstract idea.
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements are recited at a high-level of generality. It is therefore a judicial exception that is not integrated into a practical application, and does not include additional elements that are sufficient to amount to significantly more than the judicial exception. This claim is not patent eligible.
Claim 2 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea of image processing (mental process). The claim is not patent eligible.
Claim 3 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea of determining center coordinates (mental process). The claim is not patent eligible.
Claim 4 is rejected under 35 U.S.C. 101 because the claimed invention is directed to the generically recited additional element of a machine learning model, which is recited at a high level of generality such that it fails to meaningfully limit the claim. The claim is not patent eligible.
Claims 5-8 are rejected under 35 U.S.C. 101 because the claimed invention is directed to the abstract idea of determining a vector, measuring cosine similarities between vectors, and identifying a highest similarity. These all amount to mental processes, and the calculation of cosine similarity is additionally a mathematical formula. The claims are not patent eligible.
Claims 9-10 are rejected under 35 U.S.C. 101 because the claimed invention is directed to the generically recited additional elements of a masked language model and using labeled data for training. These are both recited at a high level of generality such that it fails to meaningfully limit the claim. Using masked language models and using labeled data for training are routine and highly generic. The claims are not patent eligible.
Claims 11-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a system with elements performing the abstract idea of claims 1-9. The system contains a processor, memory, and instructions. These additional elements are recited at a high level of generality such that they amount to no more than generic computer components. The claims are not patent eligible.
Claims 20 is rejected under 35 U.S.C. 101 because the claimed invention is directed to a program with instructions for performing the abstract idea of claim 1. The recited computer program product in a non-transitory computer readable medium are recited at a high level of generality such that they amount to generic computer instructions stored on generic computer components. The claim is not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-3, 11-13, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Fisher (US20140207631A1) in view of Pintsov (US20130236111A1).
Regarding claim 1, Fisher teaches “A method, comprising: receiving a document from a document provider; applying optical character recognition to the document to determine text associated with the document;” (Figure 1 shows input of invoice 102 scanned to generate image file 106, which is input to the OCR engine to generate machine-encoded text. Paragraph 23 further recites, “Invoices 102 include one or more paper invoices from one or more vendors. The invoices each include one or more billing components. In the case of telecom invoices, the billing components may represent phone numbers, circuit IDs, service charges, usage charges, usage amounts, taxes, and surcharges associated with a client's services.” Paragraphs 27-29 further recite, “Invoices 102 are processed by image scanner 104 to produce scanned image files 106. Image scanner 104 is a device that optically scans images, printed text, handwriting, or an object, and converts it to a digital image. Scanned image files 106 are digital image representations of invoices 102. … Scanned image files 106 are processed by Optical Character Recognition (OCR) engine 108. OCR engine 108 receives the scanned image files 106 and produces machine encoded text 110. … In an exemplary embodiment, OCR engine 108 produces one or more PDF files of the invoices. The PDF files contain the machine encoded text 110 generated by OCR engine 108. While the PDF file format may be used to represent machine encoded text 110, as would be appreciated by a person of skill in the art, various other file formats may be used to represent machine encoded text 110. For example, plain text files, rich text files, etc. may be used to represent machine encoded text 110.”)
While Fisher expressly discloses a processor for determining the document provider associated with the document based on the text and spatial relationships between the text (Fisher, Paragraph 35, “In order to identify the billing components and the associations between them, preprocessor 210 must first identify the vendor and billing platform associated with the invoice. In particular, the machine encoded text of the invoice is compared with a general knowledgebase that looks for any number or combinations of words and phrases, the spatial relationships of these words, and images. As would be appreciated by a person of ordinary skill in the art, various pattern matching methods may be applied to the machine encoded text in order to determine the vendor and billing platform associated with the invoice.”) Fisher does not expressly disclose determining text boxes associated with the text and corresponding coordinates, then performing the provider determination using the text boxes.
Pintsov discloses determining text boxes associated with text and corresponding coordinates, then performing document classification based on the text boxes (Pintsov, Paragraph 16, “Every document (page) image is first processed to extract text blocks (FIG. 4, step 5). This can be done by using any of the known in the art algorithms such as Docstrum method, X-Y cuts, background analysis, etc. referred to above, or by using commercially available software such as Nuance OCR, that provides in its output the text blocks (zones) found in the document images. Text blocks have two components, first being a geometric figure such as rectangle or oval or a union of such figures, and the second is the set of words contained within this geometric figure. The text blocks extracted from a document image form a collection of disjoint figures Ri which we denote T=(R1, R2, R3, . . . Rn). While coordinates are inherent to the identification of text blocks, Paragraph 17 additionally describes the use of pixel coordinates of the text blocks. Additionally, note that Figure 4 shows the use of the above text boxes and coordinates for document classification (also see Paragraphs 30-34).)
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to perform the provider determination based on text spatial properties of Fisher using text boxes (which have corresponding coordinates) generated via the teaching of Pintsov.
The motivation for doing so would have been to generate the spatial data used by Fisher. As Fisher discloses above, spatial relationships between text are used, however, Fisher is silent on the generation of the spatial data of the text. The teaching of Pintsov fills this gap. Further, one skilled in the art could have combined the elements as described above by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results. Therefore, it would have been obvious to combine Fisher with the above teaching of Pintsov to fully disclose “determining, using a processor, a plurality of text boxes associated with the text and corresponding coordinates associated with each of the text boxes; determining the document provider associated with the document based on the plurality of text boxes;”
Fisher in view of Pintsov further disclose, “and updating a database to associate the document with the document provider.” (Fisher, figure 5 shows document data with associated provider data (PSEG Electric) stored as an XML. See Figure 53, “An example hierarchical data structure is illustrated in FIG. 5. FIG. 5 shows the identified billing components from FIGS. 3 and 4 stored in a XML file. In addition to storing the billing components, the XML file captures the relationships between the various billing components. For example, billing component 340 is represented as XML element 510. Similarly, billing components 420, 425, 430, 435, 440, 450, 455, 460, and 465 are represented as XML elements 515, 520, 525, 530, 540, 545, 550, 555 and 560, respectively.”)
Regarding claim 2, Fisher in view of Pintsov disclose, “The method of claim 1,”
“wherein determining the plurality of text boxes includes performing image analysis on the document.” (Pintsov, Paragraph 16, “Every document (page) image is first processed to extract text blocks (FIG. 4, step 5). This can be done by using any of the known in the art algorithms such as Docstrum method, X-Y cuts, background analysis, etc. referred to above, or by using commercially available software such as Nuance OCR, that provides in its output the text blocks (zones) found in the document images. Text blocks have two components, first being a geometric figure such as rectangle or oval or a union of such figures, and the second is the set of words contained within this geometric figure. The text blocks extracted from a document image form a collection of disjoint figures Ri which we denote T=(R1, R2, R3, . . . Rn).” Note that this was incorporated with rationale and motivation in the rejection of claim 1.)
Regarding claim 3, Fisher in view of Pintsov disclose, “The method of claim 1,”
“wherein the corresponding coordinates associated with the text boxes are corresponding center coordinates associated with the text boxes.” (Extracted text boxes contain a plurality of pixels identified as being within the text box, including center coordinates. Therefore, center coordinates/pixel locations of the text boxes extracted (as per the rejection of claim 1) are mapped to the claimed ‘corresponding center coordinates’ as ‘the corresponding coordinates’.)
Claims 11-13 recite a system with elements corresponding to the steps recited in Claims 1-3. Therefore, the recited elements of these claims are mapped to the analogous steps in the corresponding method claims. Additionally, the rationale and motivation to combine the references apply here. Finally, the references disclose a system with a processor and memory coupled to the processor providing instructions (Fisher, Figure 7 and Paragraphs 69-70, “Computer system 700 also includes a main memory 706, preferably random access memory (RAM), and may also include a secondary memory 708. Secondary memory 708 may include, for example, a hard disk drive 710 and/or a removable storage drive 712, representing a floppy disk drive, a magnetic tape drive, an optical disk drive, or the like. Removable storage drive 712 reads from and/or writes to a removable storage unit 716 in a well-known manner. Removable storage unit 716 represents a floppy disk, magnetic tape, optical disk, or the like, which is read by and written to by removable storage drive 712. As will be appreciated by persons skilled in the relevant art(s), removable storage unit 716 includes a computer usable storage medium having stored therein computer software and/or data. In alternative implementations, secondary memory 708 may include other similar means for allowing computer programs or other instructions to be loaded into computer system 700. Such means may include, for example, a removable storage unit 718 and an interface 714. Examples of such means may include a program cartridge and cartridge interface (such as that found in video game devices), a removable memory chip (such as an EPROM, or PROM) and associated socket, a thumb drive and USB port, and other removable storage units 718 and interfaces 714 which allow software and data to be transferred from removable storage unit 718 to computer system 700.”)
Claim 20 recites a computer program product embodied in a non-transitory readable medium with instructions corresponding to the steps recited in Claim 1. Therefore, the recited programming instructions of this claim are mapped to the analogous steps in the corresponding method claim. Additionally, the rationale and motivation to combine the references apply here. Finally, the references disclose a computer program product embodied in a non-transitory readable medium (Fisher, Figure 7 and Paragraphs 69-70, “Computer system 700 also includes a main memory 706, preferably random access memory (RAM), and may also include a secondary memory 708. Secondary memory 708 may include, for example, a hard disk drive 710 and/or a removable storage drive 712, representing a floppy disk drive, a magnetic tape drive, an optical disk drive, or the like. Removable storage drive 712 reads from and/or writes to a removable storage unit 716 in a well-known manner. Removable storage unit 716 represents a floppy disk, magnetic tape, optical disk, or the like, which is read by and written to by removable storage drive 712. As will be appreciated by persons skilled in the relevant art(s), removable storage unit 716 includes a computer usable storage medium having stored therein computer software and/or data. In alternative implementations, secondary memory 708 may include other similar means for allowing computer programs or other instructions to be loaded into computer system 700. Such means may include, for example, a removable storage unit 718 and an interface 714. Examples of such means may include a program cartridge and cartridge interface (such as that found in video game devices), a removable memory chip (such as an EPROM, or PROM) and associated socket, a thumb drive and USB port, and other removable storage units 718 and interfaces 714 which allow software and data to be transferred from removable storage unit 718 to computer system 700.”)
Claim(s) 4-8, 10, and 14-18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Fisher in view of Pintsov further in view of Porter (US20200019767A1).
Regarding claim 4, Fisher in view of Pintsov disclose, “The method of claim 1,”
Fisher in view of Pintsov do not expressly disclose, “wherein determining the document provider uses a machine learning model.”
Porter teaches using a trained machine learning model for invoice class assignment (Porter. Figures 1 and 3, Paragraph 31, “In an embodiment, electronic documents may be analyzed to determine classification based on a vector comparison. Documents within a given classification may be analyzed and a unique vector determined for each document. The vector may comprise a series of floating point numbers, wherein the numbers are numeric values assigned to learned attributes of the document. The learned attributes may include features such as the layout, shape, density, position and color of the document, and other similar features. The points may form a vector having a value and direction. Documents within a given classification will have vectors of similar characteristics based on their similar features and attributes. Unclassified documents may then be processed and assigned to a classification based on comparisons between an unclassified document vector and vectors of known documents within the class. For example, the system may determine the cosine between the unclassified document vector and the known vectors within the classification. However, it will be appreciated that other comparisons may be used as well. Threshold comparison levels may be used to determine if the comparison outcome meets the classification requirements. If the threshold requirements are met then the document may be assigned to the classification and assigned to an appropriate template for data to be extracted. If the document does not meet the threshold requirements then the unclassified document vector may be compared with document vectors within a new class, or may be passed through the neural network or template comparison, as set forth above.”; Paragraphs 28-29 further describe the machine learning model used for the above: “The neural network may be trained to determine common features within a given document classification. For example, the neural network may analyze a large set of documents within a given classification to determine features that may be common to all documents within the classification. As additional documents are added within a classification, they may be used to further teach the neural network. The system may utilize the neural network to analyze unclassified documents in the queue and predict a match or likelihood of a match with a given template. Specifically, the neural network may compare the common features within the classification to features of the electronic document to determine the likelihood of a match between the document and the classification. The likelihood may be computed as a percentage confidence level of a match between the electronic document and the classification”.)
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to use a machine learning model, as taught by Porter, for the determination of document provider of Fisher in view of Pintsov.
The motivation for doing so would have been to improve classification accuracy. Further, one skilled in the art could have combined the elements as described above by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results. Therefore, it would have been obvious to combine Fisher in view of Pintsov with the above teaching of Porter to fully disclose “wherein determining the document provider uses a machine learning model.”
Regarding claim 5, Fisher in view of Pintsov further in view of Porter disclose, “The method of claim 4,”
“wherein the machine learning model represents the document provider associated with the document as a vector within a document provider space.” (Porter, Figures 1 and 3, Paragraph 31, “In an embodiment, electronic documents may be analyzed to determine classification based on a vector comparison. Documents within a given classification may be analyzed and a unique vector determined for each document. The vector may comprise a series of floating point numbers, wherein the numbers are numeric values assigned to learned attributes of the document. The learned attributes may include features such as the layout, shape, density, position and color of the document, and other similar features. The points may form a vector having a value and direction. Documents within a given classification will have vectors of similar characteristics based on their similar features and attributes. Unclassified documents may then be processed and assigned to a classification based on comparisons between an unclassified document vector and vectors of known documents within the class. For example, the system may determine the cosine between the unclassified document vector and the known vectors within the classification. However, it will be appreciated that other comparisons may be used as well. Threshold comparison levels may be used to determine if the comparison outcome meets the classification requirements. If the threshold requirements are met then the document may be assigned to the classification and assigned to an appropriate template for data to be extracted. If the document does not meet the threshold requirements then the unclassified document vector may be compared with document vectors within a new class, or may be passed through the neural network or template comparison, as set forth above.” Note that this was incorporated with motivation and rationale above. The feature space containing the above vectors is mapped to the document provider space.)
Regarding claim 6, Fisher in view of Pintsov further in view of Porter disclose, “The method of claim 5,”
“wherein a measure of similarity is computed for the vector to each document provider.” (Porter. Figures 1 and 3, Paragraph 31, “In an embodiment, electronic documents may be analyzed to determine classification based on a vector comparison. Documents within a given classification may be analyzed and a unique vector determined for each document. The vector may comprise a series of floating point numbers, wherein the numbers are numeric values assigned to learned attributes of the document. The learned attributes may include features such as the layout, shape, density, position and color of the document, and other similar features. The points may form a vector having a value and direction. Documents within a given classification will have vectors of similar characteristics based on their similar features and attributes. Unclassified documents may then be processed and assigned to a classification based on comparisons between an unclassified document vector and vectors of known documents within the class. For example, the system may determine the cosine between the unclassified document vector and the known vectors within the classification. However, it will be appreciated that other comparisons may be used as well. Threshold comparison levels may be used to determine if the comparison outcome meets the classification requirements. If the threshold requirements are met then the document may be assigned to the classification and assigned to an appropriate template for data to be extracted. If the document does not meet the threshold requirements then the unclassified document vector may be compared with document vectors within a new class, or may be passed through the neural network or template comparison, as set forth above.” Note that this was incorporated with motivation and rationale above. The “each document provider” is mapped to “vectors of known documents” within each classification.)
Regarding claim 7, Fisher in view of Pintsov further in view of Porter disclose, “The method of claim 6,”
“wherein the similarity comprises a cosine similarity.” (Porter. Figures 1 and 3, Paragraph 31, “In an embodiment, electronic documents may be analyzed to determine classification based on a vector comparison. Documents within a given classification may be analyzed and a unique vector determined for each document. The vector may comprise a series of floating point numbers, wherein the numbers are numeric values assigned to learned attributes of the document. The learned attributes may include features such as the layout, shape, density, position and color of the document, and other similar features. The points may form a vector having a value and direction. Documents within a given classification will have vectors of similar characteristics based on their similar features and attributes. Unclassified documents may then be processed and assigned to a classification based on comparisons between an unclassified document vector and vectors of known documents within the class. For example, the system may determine the cosine between the unclassified document vector and the known vectors within the classification. However, it will be appreciated that other comparisons may be used as well. Threshold comparison levels may be used to determine if the comparison outcome meets the classification requirements. If the threshold requirements are met then the document may be assigned to the classification and assigned to an appropriate template for data to be extracted. If the document does not meet the threshold requirements then the unclassified document vector may be compared with document vectors within a new class, or may be passed through the neural network or template comparison, as set forth above.” Note that this was incorporated with motivation and rationale above.)
Regarding claim 8, Fisher in view of Pintsov further in view of Porter disclose, “The method of claim 7,”
“wherein the document provider associated with the document is determined to be the document provider that is most similar to the vector within the document provider space.” (Porter. Figures 1 and 3, Paragraph 31, “In an embodiment, electronic documents may be analyzed to determine classification based on a vector comparison. Documents within a given classification may be analyzed and a unique vector determined for each document. The vector may comprise a series of floating point numbers, wherein the numbers are numeric values assigned to learned attributes of the document. The learned attributes may include features such as the layout, shape, density, position and color of the document, and other similar features. The points may form a vector having a value and direction. Documents within a given classification will have vectors of similar characteristics based on their similar features and attributes. Unclassified documents may then be processed and assigned to a classification based on comparisons between an unclassified document vector and vectors of known documents within the class. For example, the system may determine the cosine between the unclassified document vector and the known vectors within the classification. However, it will be appreciated that other comparisons may be used as well. Threshold comparison levels may be used to determine if the comparison outcome meets the classification requirements. If the threshold requirements are met then the document may be assigned to the classification and assigned to an appropriate template for data to be extracted. If the document does not meet the threshold requirements then the unclassified document vector may be compared with document vectors within a new class, or may be passed through the neural network or template comparison, as set forth above.” Note that this was incorporated with motivation and rationale above.)
Regarding claim 10, Fisher in view of Pintsov further in view of Porter disclose, “The method of claim 7,”
“wherein the machine learning model is trained using labeled data.” (Porter, Paragraphs 28-29 “The neural network may be trained to determine common features within a given document classification. For example, the neural network may analyze a large set of documents within a given classification to determine features that may be common to all documents within the classification. As additional documents are added within a classification, they may be used to further teach the neural network. The system may utilize the neural network to analyze unclassified documents in the queue and predict a match or likelihood of a match with a given template. Specifically, the neural network may compare the common features within the classification to features of the electronic document to determine the likelihood of a match between the document and the classification. The likelihood may be computed as a percentage confidence level of a match between the electronic document and the classification”. Note that the documents are analyzed for common features within a given classification, such that the documents are labeled as belonging to the classification for the training. Note that this was incorporated with motivation and rationale above.)
Claims 14-18 recite a system with elements corresponding to the steps recited in Claims 4-8. Therefore, the recited elements of these claims are mapped to the analogous steps in the corresponding method claims. Additionally, the rationale and motivation to combine the references apply here.
Claim(s) 9 and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Fisher in view of Pintsov further in view of Porter further in view of Paruchuri (US20220156300A1).
Regarding claim 9, Fisher in view of Pintsov further in view of Porter disclose, “The method of claim 4,”
Fisher in view of Pintsov further in view of Porter do not disclose “wherein the machine learning model is a masked language model.”
Paruchuri teaches document classification using a masked language model (Paruchuri, Paragraph 19, “A document classified as an unstructured document, is processed by a self-supervised model, e.g., a masked visual-language model that is pre-trained on a masked data set and subsequently trained on a task-specific data set. In an example, the masked visual language model is provided with the large pre-training data set (e.g., ˜10 million) including unlabeled documents. A predetermined small percentage of about 10% of the data in each unlabeled document in the pre-training dataset is randomly masked. The masked documents are converted into images so that each image includes a textual unit or a token with one or more words. The images are provided to the masked visual language model which detects regions of interest and predicts the masked tokens based on the contextual embedding. Besides, the masked visual language model is further trained on a smaller task-specific dataset for further fine-tuning the weights. A graphical representation of the document to be processed by the trained visual language model is initially generated. The graphical representation includes interconnected nodes wherein each node (e.g., a word) is connected to adjacent nodes with edges. The trained visual-language model processes the deep document data structure to set the weights from the text, position, and image data which is then concatenated with the node embeddings from the graphical representation of the document to generate predictions for one or more of the names/values of the entities in the tokens.”)
It would have been obvious to a person having ordinary skill in the art before the time of the effective filing date of the claimed invention of the instant application to use a masked language model, as taught by Paruchuri, for the machine learning model of Fisher in view of Pintsov further in view of Porter.
The motivation for doing so would have been to improve contextual awareness and bidirectional understanding. Further, one skilled in the art could have combined the elements as described above by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results. Therefore, it would have been obvious to combine Fisher in view of Pintsov further in view of Porter with the above teaching of Paruchuri to fully disclose “wherein the machine learning model is a masked language model.”
Claim 19 recites a system with elements corresponding to the steps recited in Claim 9. Therefore, the recited elements of this claim are mapped to the analogous steps in the corresponding method claim. Additionally, the rationale and motivation to combine the references apply here.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Maeda (US20060215878A1) teaches addressee recognition by determining candidate areas and performing matching operations.
Brdiczka (US20110258181A1) teaches estimating similarity level between documents based on extracting semantic entities
Mokhtari (US20220156488A1) teaches image- and text-based classifiers for determining classes and subclasses of input documents.
Elor (US10872236B1) teaches the classification of documents based on key-value differentiation and vector clustering.
Saito (US20200193511A1) teaches invoice classification based on cosine distances between vectors.
Park (A Methodology Combining Cosine Similarity with Classifier for Text Classification) teaches text classification by combining conventional classifiers with cosine similarity calculations for enhanced classification.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AARON JOSEPH SORRIN whose telephone number is (703)756-1565. The examiner can normally be reached Monday - Friday 9am - 5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Sumati Lefkowitz can be reached at (571) 272-3638. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/AARON JOSEPH SORRIN/
Examiner, Art Unit 2672
/SUMATI LEFKOWITZ/Supervisory Patent Examiner, Art Unit 2672
