DETAILED ACTION
This Office Action is responsive to amendments and arguments filed on December 8th, 2025. Claims 1, 8 and 15 are amended, claims 1-20 are pending; hence, this Action has been made FINAL.
Any objections/rejections not mentioned in this Office Action have been withdrawn by the Examiner.
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 .
Response to Amendments and Arguments
With respect to rejection made under 35 U.S.C. 101, Applicant argues, “claim 1 recites one or more elements that cannot be practically performed in the human mind… the method requires generating predictions via a multi-modal encoder-based transformer which includes a plurality of encoders, an ensemble layer, and a task head. The human mind is not equipped to perform at least these claim limitations. For these reasons alone, Applicant submits that the claims are patent eligible under Prong One and no further inquiry is needed under Prong Two,” (beginning on page 10 of Remarks).
Applicant’s argument is persuasive; accordingly, the rejections under 35 U.S.C. 101 are withdrawn.
With respect to rejections made under 35 U.S.C. 103, Applicant argues, “the combination of Drillock, Rimchala, and Karpets fails to disclose, teach, or suggest every limitation of the present claims… Drillock fails to disclose for each token position in the document: generating, by the multi-modal encoder-based transformer model, a plurality of probability predictions across overlapping context windows of the document, and computing, from the plurality of probability predictions, a joint probability that the token is an entity.” (beginning on page 13 of Remarks).
Applicant further argues, “Rimchala extracts the entity information by determining an entity value and associating the entity value with an entity identifier, forming an entity value and entity identifier pair. Rimchala ¶ [0012]… In contrast, the present claims recite for each token position in the document: generating, by the multi-modal encoder-based transformer model, a plurality of probability predictions across overlapping context windows of the document, and computing, from the plurality of probability predictions, a joint probability that the token is an entity,” (page 14 of Remarks) and, “Karpets is generally directed to automatic legal decision making and optimizing legal activities with specified criteria. Karpets, Abstract. Applicant respectfully submits that Karpets, cited as allegedly teaching generating a summary of the document, fails to cure the deficiencies of Drillock and Rimchala,” (page 14 of Remarks).
Applicant’s arguments are moot, as the amended claims fall within the teachings of Rimchala. The rejections under 35 U.S.C. 103 are maintained. Further details are provided below.
Claim Rejections - 35 USC § 103
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.
Claims 1, 6, 8, 13, 15 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Patent 12,332,911 to Drillock et al. (hereinafter, "Drillock") in view of European Patent Application EP 4220575 to Rimchala and Frick (hereinafter, "Rimchala"), and further in view of International Publication WO 2021/112704 to Karpets et al. (hereinafter, "Karpets").
Regarding claims 1, 8 and 15, Drillock teaches a system, computer-readable medium and method “of processing a document, comprising: receiving, by a computing system, a document to be analyzed, the document associated with a document type of a plurality of document types; (column 6, lines 42-46, "In one or more implementations, a document is received, such as an attachment in an email message from a subscribing company, and the document is processed to identify, extract, and generate data therefrom, for example, for mapping to a respective schema.")determining, by the computing system, the document type associated with the document; (column 7, line 60, through column 8, line 5, "In respective implementations, features of the present disclosure, such as segmentation and selections of a respective model can be disabled or enabled. In one or more implementations, a parser can be configured with an appropriate classification type (e.g., a contract document or an invoice document). An example create new project display screen 1700 is shown in FIG. 17, including NLP section 1702 identifying in which an INVOICES category and CPTY CONTRACTS category have been created and configured for respective the parsers. In the event a document 304 is imported into an uncategorized document type, one or more auto-classification procedures can be implemented.")routing, by the computing system, the document to a plurality of name entity recognition transformer models trained to identify a plurality of entities in the document; (column 12, lines 16-23, "Still further, the parsing pipeline of the present disclosure can support segmentation for complex documents that include a plurality section groups, sections, and/or subsections. Still further, entity recognition (also referred as “named entity recognition”) can be applied against virtually any portion of a document, and a plurality of targeted models can be applied against the document, section group, section, and/or subsection.").
Drillock does not explicitly teach “extracting, by the plurality of name entity recognition transformer models, token-level outputs representing the plurality of entities from the document ,” “providing, by the computing system, the token-level outputs together with word embeddings of the document to a multi-modal encoder-based transformer model ,” or “generating, by the multi-modal encoder-based transformer model, a plurality of probability predictions across overlapping context windows of the document, and computing, from the plurality of probability predictions, a joint probability that the token is an entity,” and thus, Rimchala is introduced.
Rimchala teaches “extracting, by the plurality of name entity recognition transformer models, token-level outputs representing the plurality of entities from the document;” (paragraph [0022], "The OCR engine may identify text as a linear sequence of individual characters with corresponding character bounding boxes from the document. Then, a tokenizer partitions the linear sequence of characters into sub-word pieces, words or terms. The tokenization process may convert the characters into token values. The token values may be useful semantic units for further processing.").
Rimchala further teaches “providing, by the computing system, the token-level outputs together with word embeddings of the document to a multi-modal encoder-based transformer model (paragraph [0026], "Turning to the machine learning framework, the machine learning framework implements a layoutLMv2 model having a repeated stack of transformer's encoder layers each of which encodes bidirectional representations from the multi-modal inputs (e.g., text, layout, and image) in one or more embodiments… The input to the machine learning framework is the text and layout (212) and the image feature vector (214) in or more embodiments. The text and layout is input to an embedding process that converts the token values into word embeddings."), the multi-modal encoder-based transformer model comprising:a plurality of encoders (paragraph [0029], “The encoder portion of the transformer model may include a stack of encoders.”), each encoder configured to process the token-level outputs and the word embeddings (paragraph [0027], “The input to the machine learning framework is the text and layout (212) and the image feature vector (214) in or more embodiments. The text and layout is input to an embedding process that converts the token values into word embeddings.”), an ensemble layer configured to combine encoder outputs (paragraph [0029], “An output of a first encoder layer in the encoders may be an input of a second encoder layer.”), anda task head configured to produce entity probabilities;” (paragraph [0046], "The confidence head (220) assigns a probability of exact match prediction to each predicted entity.")Rimchala furter teaches “for each token position in the document;generating, by the multi-modal encoder-based transformer model, a plurality of probability predictions across overlapping context windows of the document (paragraph [0037], “The token extraction head (218) classifies each token value into the predefined set of token labels by identifying the corresponding probability that the token value belongs to each token label. In one or more embodiments, the predefined set includes the superset of token labels for the various document types processed by the system. Thus, the output of the token extraction head (218) is a set of probabilities, whereby the token value is assigned a token label of the set having the greatest probability.”), andcomputing, from the plurality of probability predictions, a joint probability that the token is an entity;“ (paragraph [0039], "The classified tokens (228) are passed to a token aggregator (230). The token aggregator (230) is configured to aggregate tokens into entity values using a set of rules applied over the sequence of predicted tokens classifier outputs.").
Drillock and Rimchala are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock with the teachings of Rimchala for the purpose of improving entity recognition accuracy. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
The combination of Drillock and Rimchala does not teach “generating, by the computing system, a summary of the document by arranging the plurality of entities in accordance with an ontology dedicated to the document type,” and thus, Karpets is introduced. Karpets teaches “generating, by the computing system, a summary of the document by arranging the plurality of entities in accordance with an ontology dedicated to the document type;” (paragraph [0024], "In another preferred embodiment of the claimed technical solution, an automatic legal decision-making system is provided, comprising: an intelligent document recognition system including a processing module that receives a request for a legal decision; obtaining at least one document associated with said request; an entity extraction module, which ensures that at least one entity related to the request for a legal decision is identified in the document and extracted from the document using a machine learning algorithm; an automated decision-making system that provides legal analysis of said at least one extracted entity using a set of rules generated using a domain-specific language; formation of a legal decision on the basis of a legal analysis of said extracted entities of said at least one document.").
Drillock, Rimchala and Karpets are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock and Rimchala with the teachings of Karpets for the purpose of providing a usable output from document processing. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
Regarding claims 6, 13 and 19, Karpets further teaches “generating, by the computing system, the summary of the document by arranging the plurality of entities in accordance with the ontology dedicated to the document type comprises: performing a recovery process on the plurality of entities by determining whether any tokens were excluded or incorrectly included in a group of entities;” (paragraphs [0025] and [0026], "In one particular implementation example, the system further comprises a retrieved entity verification module that checks the correctness of the retrieved attributes; In another particular example of the implementation of the system, the verification module analyzes the semantic correspondence of at least one extracted entity to the portion of the text of the document from which the extraction was performed.").
Drillock, Rimchala and Karpets are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock and Rimchala with the teachings of Karpets for the purpose of improving entity recognition accuracy. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
Claims 2-5, 7, 9-12, 14, 16-18 and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Drillock, Rimchala and Karpets as applied to claims 1, 8 and 15 above, and further in view of International Publication WO 3033/071917 to Narayanan and Shi (hereinafter, "Narayanan").
Regarding claims 2, 9 and 16, the combination of Drillock, Rimchala and Karpets does not teach “causing, by the computing system, display of the summary of the document,” however, Narayanan teaches “causing, by the computing system, display of the summary of the document;” (paragraph [023], "The first NLP model may be configured to generate a first mark-up that represents aspects of the text document at a relatively high level of generality. For example, the first mark-up may include sentences summarizing the text document as a whole, a listing of entities detected within the text document, categories of text identified within the text document, and/or a representation of a paragraph structure of the text document. The first mark-up may be displayed, along with the text document, by way of a user interface, which may allow the reader to interact with portions of the first mark-up.").
Drillock, Rimchala, Karpets and Narayanan are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock, Rimchala and Karpets with the teachings of Narayanan for the purpose of improving entity recognition usability. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
Regarding claims 3 and 10, Narayanan further teaches “display of the summary of the document further comprises: causing display of the document side-by-side with the summary of the document;” (paragraph [023], "The first NLP model may be configured to generate a first mark-up that represents aspects of the text document at a relatively high level of generality. For example, the first mark-up may include sentences summarizing the text document as a whole, a listing of entities detected within the text document, categories of text identified within the text document, and/or a representation of a paragraph structure of the text document. The first mark-up may be displayed, along with the text document, by way of a user interface, which may allow the reader to interact with portions of the first mark-up.").
Drillock, Rimchala, Karpets and Narayanan are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock, Rimchala and Karpets with the teachings of Narayanan for the purpose of improving entity recognition usability. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
Regarding claims 4, 11 and 17, the combination of Drillock, Rimchala and Karpets does not teach “generating, by the computing system, the summary of the document by arranging the plurality of entities in accordance with the ontology dedicated to the document type comprises: grouping entities in the document based on primitive type and structure defined by the ontology,” however, Narayanan teaches “grouping entities in the document based on primitive type and structure defined by the ontology;” (paragraph [069], "In a first example, first NLP theme model 350 may be a coarse text classification model configured to generate a plurality of categories of text present within text document 300. Text classification models may include word, phrase, and/or sentence embedding algorithms, and/or algorithms configured to cluster the embeddings generated thereby.").
Drillock, Rimchala, Karpets and Narayanan are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock, Rimchala and Karpets with the teachings of Narayanan for the purpose of improving entity recognition usability. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
Regarding claims 5, 12 and 18, Narayanan further teaches “determining whether a sequence of entities is associated or related; (paragraph [070], "The second NLP theme model 354 may be a refined text classification model configured to identify, within text document 300, one or more sentences that are associated with a particular category. Thus, for example, the user may be provided with a list of categories associated with text document 300 and, based on selecting a particular category, the user may be provided with an indication of the one or more sentences in text document 300 that belong to the particular category.") and grouping the sequence of entities under a single identifier;” (paragraph [070], "The second NLP theme model 354 may be a refined text classification model configured to identify, within text document 300, one or more sentences that are associated with a particular category. Thus, for example, the user may be provided with a list of categories associated with text document 300 and, based on selecting a particular category, the user may be provided with an indication of the one or more sentences in text document 300 that belong to the particular category.").
Drillock, Rimchala, Karpets and Narayanan are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock, Rimchala and Karpets with the teachings of Narayanan for the purpose of improving entity recognition usability. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
Regarding claims 7, 14 and 20, the combination of Drillock, Rimchala and Karpets does not teach “generating, by the computing system, the summary of the document by arranging the plurality of entities in accordance with the ontology dedicated to the document type comprises: assembling phrases in accordance with a primitive type assigned to labels of tokens associated with the plurality of entities,” however, Narayanan teaches “assembling phrases in accordance with a primitive type assigned to labels of tokens associated with the plurality of entities;” (paragraphs [084] and [085], "Figure 6 illustrates example mark-up generated by NLP theme models (e.g., NLP theme models 350 and 354). Specifically, Figure 6 illustrates GUI 600 configured to display therein text document 604, theme mark-up 602, and theme mark-up 608. Specifically, text document 604 may be processed by a text classification model to generate theme mark-up 602, which includes bullish outlook/category 610, neutral outlook/category 612, and bearish outlook/category 614. Depending on the content of text document 604 and/or the type of text classification model used, theme mark-up 602 may include a plurality of different categories.").
Drillock, Rimchala, Karpets and Narayanan are considered analogous because they are each concerned with document processing and entity recognition. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to have modified Drillock, Rimchala and Karpets with the teachings of Narayanan for the purpose of improving entity recognition usability. Given that all the claimed elements were known in the prior art, one skilled in the art could have combined the elements by known methods with no change in their respective functions, and the combination would have yielded nothing more than predictable results.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
U.S. Patent 12,260,177 to Dauhiala et al.
U.S. Patent 8,429,099 to Perkowitz et al.
U.S. Patent 8,788,523 to Martin et al.
U.S. Patent Application Publication 2021/0374348 to Dasgupta et al.
U.S. Patent Application Publication 2021/0390256 to Liu et al.
U.S. Patent Application Publication 2024/0070794 to Bonfante et al.
U.S. Patent Application Publication 2024/0160953 to Manda et al.
U.S. Patent Application Publication 2024/0370479 to Hudetz et al.
U.S. Patent Application Publication 2024/0394293 to Lauber.
Australia Patent Application AU 2019203783 to Dasgupta et al.
China Invention Application CN 115129870 to Zuo et al.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SEAN T SMITH whose telephone number is (571)272-6643. The examiner can normally be reached Monday - Friday 8:00am - 5:00pm.
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, PIERRE-LOUIS DESIR can be reached at (571) 272-7799. 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.
/SEAN THOMAS SMITH/Examiner, Art Unit 2659
/PIERRE LOUIS DESIR/Supervisory Patent Examiner, Art Unit 2659