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 .
Receipt of Applicant’s Amendment filed July 7, 2026, is acknowledged.
Response to Amendment
Claims 1, 10, and 16 have been amended. Claims 2, 3, 11, and 12 have been canceled. Claims 19-24 are new. Claims 1, 4-10, and 13-24 are pending and are provided to be examined upon their merits.
Response to Arguments
Applicant's arguments filed July 7, 2026, have been fully considered but they are not persuasive. A response is provided below in bold where appropriate.
Applicant argues 35 USC §101 Rejection, starting pg. 12 of Remarks:
Rejection under 35 U.S.C. $101
Claims 1, 4-10, and 13-24 stand rejected under 35 U.S.C. § 101 as allegedly being directed to non-statutory subject matter. For at least the reasons discussed during the interview and without acquiescing in the rejections of the outstanding Office Action, Applicant respectfully traverses this rejection. As amended, claim 1 and claim 10, and the claims that depend therefrom, represent patent-eligible subject matter under 35 U.S.C. § 101 because the claims fall within statutory process and system categories, do not recite a judicial exception, and integrate the claimed features into a practical application that improves the functioning of natural language processing computers and systems.
Step 2A - The Claims Are Not "Directed To" an Abstract Idea
Under MPEP § 2106.04, a two-prong approach is used to determine whether a claim is directed to a patent-ineligible concept in Step 2A. "Examiners determine in Prong One whether a claim recites a judicial exception, and if so, then determine in Prong Two if the recited judicial exception is integrated into a practical application of that exception. Together, these prongs represent the first part of the Alice/Mayo test, which determines whether a claim is directed to a judicial exception." See MPEP § 2106.04, subsection II.A. Only if the claim recites a judicial exception under Prong One and fails to integrate the judicial exception into a practical application under Prong Two will a claim be considered to be directed to the judicial exception (e.g., an abstract idea) under Step 2A of the eligibility analysis.
Step 2A Prong One-The Claims Do Not Recite a Judicial Exception
During the assessment of claims 1 and 10 under Step 2A's first prong, the Examiner asserts that the claims involve judicial exceptions tied to mathematical concepts, organizing human activity, and mental processes, as detailed in the Office Action on pages 3 and 4. Applicant respectfully disagrees.
As per MPEP § 2106.04(a), abstract ideas fall into three groups: 1) mathematical concepts, which involve mathematical relationships, formulas, equations, and calculations; 2) specific methods of organizing human activity, which cover basic economic principles, commercial or legal interactions, and personal behavior or relationships management; and 3) mental processes, which encompass concepts executed within the human mind, such as observations and judgments. MPEP § 2106.04(a) also notes that if identified elements do not fit these groups, concluding that the claim is not directed to an abstract idea is reasonable.
For the mathematical concepts subgroup, claim 1 and claim 10 do not recite a mathematical concept, equation, or formula. Rather, the claims recite specific computational data structures and memory-bound operations. Specifically, claim 1 and claim 10 recite: "generate, using a pretrained language model, a tensor representation of at least a portion of the transcript that includes the two or more transcript concepts of the concept combination, wherein the tensor representation is stored in a computer memory and has dimensions of token by dims, wherein dims represents a length of word vectors of the pretrained language model." Specifically, a tensor representation, as recited in claim 1 and claim 10, is a multidimensional array data structure stored within physical computer memory. This structure maps high- dimensional vector spaces representing token embeddings and is defined by specific hardware- bound dimensions. Slicing this multidimensional numerical array in memory at specific token coordinate offsets and concatenating the resulting slices are technical, computer-implemented data-manipulation operations performed on memory structures to capture precise contextual properties. The claims do not recite, nor do they preempt, any mathematical formula, equation, or algorithm.
The tensor/vector is a mathematical representation of text (transcript concepts). Generating tensor to represent text is creating a mathematical form of text which is a mathematical concept.
For the subgroup "certain methods of organizing human activity," the Examiner asserts that because the claims recite a conversation between a physician and a patient, the features cover managing interactions or relationships between people. Applicant respectfully traverses this assertion. The claims do not involve actions such as marketing, sales behaviors, or business relations. This subgroup is confined to economic, commercial, legal, and interpersonal activities. Claim 1 and claim 10 are directed to an automated computer-implemented data processing and documentation tool, which belongs to the technical field of natural language processing. The conversation between the physician and the patient serves as a passive environmental context and a source of digital input data for the system. The claims do not govern, manage, or organize human behavior or interactions. As the Court of Appeals for the Federal Circuit recites in Ollnova Technologies Ltd. v. ecobee Technologies ULC, 2025-1045, opinion at 30 (Fed. Cir. June 4, 2026), human-driven analogies fail when the analogies fail to account for the technical challenges addressed by the inventors and the claimed improvements to system operation, noting that "[i]t is not enough, however, to merely trace the invention to some real-world analogy." Because the claims do not manage or organize personal relationships, but rather process digital data inputs, the claims do not recite a method of organizing human activity. Even if a doctor- patient conversation were deemed to fall within one of the enumerated categories, characterizing the claims as reciting "certain methods of organizing human activity" would still be improper. The relevant inquiry is whether the claims are "directed to" an abstract idea. Here, claims 1 and 10 plainly are not directed to any method of organizing human activity. Rather, as explained above, the claims are directed to an automated, computer-implemented data processing and documentation tool in the technical field of natural language processing.
The requirement is to look at elements of the claim to determine if they recite abstract concepts.
From Applicant’s specification…
“In recent years, extensive research and development effort in healthcare industry has been focused on automating a process of generating the clinical note based on doctor-patient conversations (audio or transcript). Typically, in case of doctor-patient conversation transcripts, pieces of information (e.g., words, phrases, etc.) may need to be extracted from the transcript. Such pieces of information may be combined with other pieces to form a target structed data. Structured data is data organized into specific fields (or categories) as part of a schema, with each field having a defined purpose. Examples of such fields may include numbers, anatomical structures, laterality, etc. Structured data categories may include patient history, family history past surgeries, medications, exam results, vital signs, and more.” (pg. 1, lines 14-22)
Therefore, conversations and notes between doctor-patient falls under certain methods of organizing human activity.
Claim 1 and claim 10 also do not recite a mental process. The mental processes grouping is limited to concepts that can practically be performed in the human mind, such as observation, evaluation, judgment, or opinion. As explained in the USPTO Memorandum, Reminders on Evaluating Subject Matter Eligibility of Claims Under 35 U.S.C. § 101 (Aug. 4, 2025), claim features that encompass AI "in a way that cannot be practically performed in the human mind do not fall within this grouping." Here, the claims require computer-implemented operations performed on electronic data structures in memory, including generating a tensor representation using a pretrained language model, slicing that tensor at specified token positions, and concatenating resulting vectors into a machine-usable representation. These are not acts that a human can practically perform mentally or with pen and paper. Rather, they are computer- centric operations carried out on digital data arranged and processed in computer memory. As further described in paragraph [0082] of the specification, the disclosed techniques are outside the realm of what is practicable for a human to perform and are inextricably linked to computer implementation. Accordingly, claim 1 and claim 10 do not recite a mental process.
Respectfully, most if not all of the steps can be performed in the mind of a person and with pen and paper.
Since claim 1 and claim 10 do not recite an abstract idea, claim 1 and claim 10 are eligible for patenting.
For at least the foregoing reasons, amended claim 1 and claim 10, along with their dependent claims, are patent-eligible under Step 2A's Prong One. Applicant respectfully asserts that the rejections of claims 1, 4-10, and 13-24 under 35 U.S.C. § 101 should be withdrawn.
Step 2A Prong Two-If the Claim Recites a Judicial Exception, the Judicial Exception Is Integrated into a Practical Application
Even if the claim were determined to recite a judicial exception, which is not the case here, a claim is patent-eligible under Step 2A Prong Two if the claim as a whole integrates the recited judicial exception into a practical application of that exception. See MPEP § 2106.04(d)(I). A claim integrates a judicial exception into a practical application when the additional features provide a technical solution to a technical problem or improve the functioning of a computer or other technology.
In the instance of claim 1 and claim 10, the entirety of the claim integrates the alleged abstract idea into a practical, computer-implemented clinical charting application. As described in paragraph [0003] of the application publication, automated extraction of structured data from patient-physician conversations faces technical hurdles, including unstructured conversational noise, vocabulary variation, and the fact that related medical concepts are often mentioned far apart across a transcript. The claimed features provide a specific, automated clinical documentation tool that solves these computer-centric natural language processing challenges.
A combination of abstract elements is still abstract.
The Examiner asserts that the computer hardware and machine learning model are recited at a high level of generality, and that machine learning can be performed with pen and paper. Applicant respectfully traverses this assertion. In the USPTO Appeals Review Panel decision Ex parte Desjardins, Appeal 2024-000567, opinion at 9 (PTAB Sept. 26, 2025), the Panel addressed evaluating machine learning claims at a high level of generality, reciting that categorically excluding artificial intelligence innovations by treating machine learning as a generic algorithm on conventional computer components is legally improper and fails to account for the specific system configurations. Just as in Desjardins, the current claims reflect a specific, non-generic computational sequence that improves how the machine learning model itself operates.
Desjardins improved AI technology. Respectfully, Applicant is not claiming AI innovations as in Desjardins.
Specifically, instead of requiring the machine learning model to ingest and process a sparse, high-dimensional tensor of an entire transcript, the claimed system generates a tensor representation using a pretrained language model, determines specific slices of that tensor representation corresponding to precise token positions of transcript concepts, and concatenates those slices to form a compact, dense concatenated tensor representing only the concept combination. This targeted slicing and concatenation scheme creates a compressed, computationally efficient input vector. By presenting the machine learning classifier with this optimized, dense input structure, the computational overhead of the classification model is minimized. This improves the functioning of the natural language processing computer by reducing CPU instruction cycles and optimizing RAM utilization during model inference.
Respectfully, the above reducing CPU cycles and optimizing RAM utilization is an effect or result of an abstract process of generating and using a mathematical construct. Computer technology itself is not improved.
Furthermore, as recited in dependent claim 20, the system is configured to filter the transcript by discarding portions of the transcript that do not contain any of the plurality of transcript concepts prior to generating the tensor representation, thereby reducing a computational load of the pretrained language model and the machine learning model relative to processing an entirety of the transcript. As the Court of Appeals for the Federal Circuit recites in Ollnova Technologies Ltd. v. ecobee Technologies ULC, 2025-1045, opinion at 29 (Fed. Cir. June 4, 2026), claims are patent-eligible under Step 2A when they recite "a particularized set of constraints that alters when and how data is collected and transmitted" representing a "particular means... rather than a generalized data-handling concept." The transcript filtering and specialized tensor slicing of the claimed system represent a particularized set of data-processing constraints that directly reduce the processing load on the computer processor, enabling the natural language processing system to satisfy the temporal constraints of real-time clinical charting.
From MPEP 2106.04(a)…
“To facilitate examination, the Office has set forth an approach to identifying abstract ideas that distills the relevant case law into enumerated groupings of abstract ideas. The enumerated groupings are firmly rooted in Supreme Court precedent as well as Federal Circuit decisions interpreting that precedent, as is explained in MPEP § 2106.04(a)(2) (s2106.html#ch2100_d29a1b_13ae3_321) . This approach represents a shift from the former case-comparison approach that required examiners to rely on individual judicial cases when determining whether a claim recites an abstract idea. By grouping the abstract ideas, the examiners’ focus has been shifted from relying on individual cases to generally applying the wide body of caselaw spanning all technologies and claim types.
Therefore, the MPEP and not specific cases are relied upon in the analysis. Reducing processing load is an effect and not an improvement to computer technology itself.
The above cited case had non-abstract steps that in combination provided more than a generalized data handling concept of polling and selective transmission as a particular communication means (pg. 29, para. 2). The specification cites a technical benefit (pg. 9, para. 2) of reduced bandwidth usage and delays. Applicant’s steps are abstract as they can be performed as a mental process and are not improving a technology itself.
This specific technical optimization enables the system to satisfy the real-time processing constraints necessary for clinical charting. Claim 1 and claim 10 recite: "wherein the output structured data is generated before the physician completes charting on the patient." Paragraph [0084] of the application publication recites that clinical documentation systems achieve high throughput to complete operations before the doctor completes charting on a given patient, noting that because of the large volume of data, it would take a human many weeks or months to perform these steps, making the approach without a machine learning model impracticable. Generating the output structured data within this specific temporal constraint is made computationally possible by the optimized filtering, slicing, and concatenation operations. This represents a concrete improvement in the functioning of a computer-implemented natural language processing system. The August 4, 2025 USPTO Reminders Memo recites: "The specification does not need to explicitly set forth the improvement, but it must describe the invention such that the improvement would be apparent to one of ordinary skill in the art. The claim itself does not need to explicitly recite the improvement described in the specification."
Automating a manual process has not been found to be enough to make abstract claims statutory.
From MPEP 2106.05(a) I…
“Examples that the courts have indicated may not be sufficient to show an improvement in computer-functionality:…
…iii. Mere automation of manual processes, such as using a generic computer to process an application for financing a purchase, Credit Acceptance Corp. v. Westlake Services, 859 F.3d 1044, 1055, 123 USPQ2d 1100, 1108-09 (Fed. Cir.2017) or speeding up a loan-application process by enabling borrowers to avoid physically going to or calling each lender and filling out a loan application, LendingTree, LLC v. Zillow, Inc., 656 Fed. App'x 991, 996-97 (Fed. Cir. 2016) (non-precedential);…”
In addition, as recited in dependent claims 8, 9, and 19, the system is configured to link the output structured data back to the corresponding transcript concepts in the transcript and visually highlight and graphically link them within a user interface. This provides a concrete, computer-implemented evidentiary audit trail and visual interface that is a specific technical feature of the documentation tool, further integrating any alleged exception into a practical clinical charting application.
The above can be done with pen and paper.
In Finjan Inc. v. Blue Coat Systems, Inc., 879 F.3d 1299 (Fed. Cir. 2018), the court found a system eligible because the claims recited a specific, non-generic utility (improved security profiles) that improved the computer technology. Similarly, claim 1 and claim 10 integrate the data processing steps into a specialized clinical documentation utility that optimizes computer resource usage and satisfies temporal charting constraints, consistent with the technological integration in Finjan.
Finjan improved computer security. Applicants are not claiming an improvement to computer security itself.
For at least these reasons, claim 1 and claim 10 as a whole integrate any alleged abstract idea into a practical application. Therefore, the claims are not directed to a judicial exception and are eligible under 35 U.S.C. § 101.
The Examiner respectfully maintains the claims are abstract.
Step 2B-The Claims Provide an Inventive Concept
Even if the Office assumes, contrary to the position of Applicant, that claim 1 and claim 10 lack integration into a practical application, the claims are eligible under Step 2B. Under Step 2B, the Office is to assess whether an element or combination of elements in the claim provides an inventive concept, which involves evaluating whether the features offer significantly more than the exception.
In claim 1 and claim 10, there exist specific features and combinations thereof that go beyond routine, conventional activity in the field. This establishes an inventive concept, thereby rendering the claims eligible. These elements connect the disclosed technology to a technical context that extends beyond the judicial exception itself.
The ordered combination of filtering a transcript, generating a dimensional tensor representation in computer memory, slicing the tensor representation at specific token positions, concatenating the slices, and classifying the concatenated tensor via a machine learning model to generate structured data before the physician completes charting is not well-understood, routine, or conventional in the field. The applied references do not disclose this non-conventional, technically optimized combination of elements. Under Berkheimer v. HP Inc., 881 F.3d 1360, 1369 (Fed. Cir. 2018), whether a combination of elements is well-understood, routine, or conventional is a question of fact that is required to be supported by evidence. As set forth in the Berkheimer memorandum, an additional element or combination of elements should not be deemed conventional unless supported by a citation to a statement in the specification, a citation to a court decision, a citation to a publication, or a statement of official notice. The Office Action fails to provide any such evidentiary support.
Applicant has amended their claims to include tensor.
The Examiner did not recite well-understood, routine and conventional it the rejection.
From MPEP 2106.07(a) III…
“When performing the analysis at Step 2A Prong One, it is sufficient for the examiner to provide a reasoned rationale that identifies the judicial exception recited in the claim and explains why it is considered a judicial exception (e.g., that the claim limitation(s)falls within one of the abstract idea groupings). Therefore, there is no requirement for the examiner to rely on evidence, such as publications or an affidavit or declaration under 37 CFR 1.104(d)(2) (mpep-9020-appx-r.html#d0e322249), to find that a claim recites a judicial exception. Cf. Affinity Labs of Tex., LLC v. Amazon.com Inc., 838 F.3d 1266, 1271-72, 120 USPQ2d 1210, 1214-15 (Fed. Cir. 2016) (affirming district court decision that identified an abstract idea in the claims without relying on evidence); OIP Techs., Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1362-64, 115 USPQ2d 1090, 1092-94 (Fed. Cir. 2015) (same); Content Extraction &Transmission LLC v. Wells Fargo Bank, N.A., 776 F.3d 1343, 1347, 113 USPQ2d 1354, 1357-58 (Fed. Cir. 2014) (same).
At Step 2A Prong Two or Step 2B, there is no requirement for evidence to support a finding that the exception is not integrated into a practical application or that the additional elements do not amount to significantly more than the exception unless the examiner asserts that additional limitations are well-understood, routine, conventional activities in Step 2B.”
Furthermore, the specialized data-processing architecture recited in the claims cannot be dismissed as generic data-handling. In the Appeals Review Panel decision Ex parte Desjardins, Appeal 2024-000567, opinion at 9 (PTAB Sept. 26, 2025), the Panel addressed evaluating machine learning claims at a high level of generality, reciting that evaluating artificial intelligence innovations as a generic algorithm on conventional computer components is legally improper and jeopardizes technological leadership. Just as in Desjardins, the current claims recite a specific, technically optimized combination of operations that improves the functioning of the natural language processing system.
Respectfully, there is no indication Applicant has improved AI technology itself.
This technical focus distinguishes the present claims from those held ineligible in Technology in Ariscale, LLC v. Razer USA Ltd., 24-1657 (Fed. Cir. Jan. 6, 2026). In Technology in Ariscale, the Court of Appeals for the Federal Circuit found the claims ineligible because the claims failed to capture the specific deinterleaving and combining schemes described in the specification. In contrast, the present claims recite the specific technical features (generating a tensor representation in computer memory, slicing the representation at specific token positions to obtain contextual vectors, and concatenating the slices), capturing the specific structural and functional details that achieve the technical improvements described in paragraph [0084] of the specification of the application publication. Because the claims recite a non- conventional, technically optimized combination of elements that improves natural language processing, the claims provide an inventive concept.
Respectfully, there is no indication Applicant is improving natural language processing.
For at least the foregoing reasons, claims 1, 4-10, and 13-24 recite patent-eligible subject matter under 35 U.S.C. § 101. Applicant respectfully requests that the Examiner reconsider and withdraw the rejection under 35 U.S.C. § 101.
Based on the above response, the rejection is respectfully maintained.
Applicant argues 35 USC §112(b) Rejection, starting pg. 21 of Remarks:
Rejection under 35 U.S.C. $ 112(b)
Claims 1 and 4-9 stand rejected under 35 U.S.C. § 112(b). Without acquiescing in this rejection, but merely to expedite prosecution, Applicant amends claim 1 to address the Examiner's concerns. Thus, claims 1 and 4-9 are definite under 35 U.S.C. § 112(b).
Accordingly, Applicant respectfully requests that the Examiner reconsider and withdraw the rejection of claims 1 and 4-9 under 35 U.S.C. § 112(b).
Withdrawn based on the claim amendments.
Applicant argues 35 USC §103 Rejection, starting pg. 22 of Remarks:
Rejection under 35 U.S.C. $ 103 based on LEVENTHAL and MARTIN
Claims 1, 4-10, and 13-24 stand rejected under 35 U.S.C. § 103 as allegedly being unpatentable over LEVENTHAL in view of MARTIN and further in view of THOMPSON W. Applicant respectfully traverses this rejection. The cited sections of the applied references, whether taken alone or in any combination, do not disclose the specific computational data- processing architecture recited in the amended claims.
Claim 1 recites, in part, configured to: "generate, using a pretrained language model, a tensor representation of at least a portion of the transcript . .. determine a plurality of slices of the tensor representation, wherein each slice of the plurality of slices corresponds to a token position of a respective one of the two or more transcript concepts . .. to obtain a vector of length dims containing semantic and contextual properties of the respective transcript concept, and concatenate the plurality of slices to form a concatenated tensor representing the concept combination." In contrast, THOMPSON W does not disclose slicing a generated multidimensional tensor representation at specific, identified token positions of concepts. Rather, THOMPSON W performs a raw text-splitting or segmenting operation on unstructured text before any model encoding to generate generic text snippets or segments. Specifically, paragraph [0084] of THOMPSON W recites: "Accordingly, some embodiments segment or split the text (e.g., the unstructured text corresponding to each episode) into roughly even snippets 226, taking sentence boundaries into account." Additionally, paragraph [0085] of THOMPSON W recites: "This list is segmented (410) or split into M segments 412, each segment having a predetermined number of tokens (256 in this example)." This raw segmenting of unstructured text into generic segment lengths is structurally and functionally distinct from the claimed operation of first generating a multidimensional tensor representation of a transcript in a computer memory using a pretrained language model, and then slicing that generated tensor representation at specific, identified token positions of concept combinations to extract contextual vectors of length dims.
Respectfully, Applicant above is arguing things not claimed. Applicant is also claiming things not taught.
From Applicant’s argument above…
>>”In contrast, THOMPSON W does not disclose slicing a generated multidimensional tensor representation at specific, identified token positions of concepts. Rather, THOMPSON W performs a raw text-splitting or segmenting operation on unstructured text before any model encoding to generate generic text snippets or segments.”<<
From Applicant’s claim 10…
“determining, via the at least one computer processor, a plurality of slices of the tensor representation, wherein each slice of the plurality of slices corresponds to a token position of a respective one of the two or more transcript concepts of the concept combination…”
Therefore, slicing and identified positions are not claimed.
From Applicant’s specification on token…
“As used herein, the term "tensor" generally refers to a linear geometrical quantity. In the present description, the tensor is typically in the dimension of "[token, dims]", where "dims" is a length of word vectors that the text encoder (e.g., the text encoder 162) in use employs. Creating a slice means taking a portion of the tensor at a token ID corresponding to the transcript concept and obtaining the [dims] vector. This contains semantic and contextual information about a specific mention of the transcript concept.” (pg. 14, lines 19-24)
Therefore, a token ID corresponds to a transcript concept, not a token position. The “token position” is not taught in the disclosure.
A slice is created by taking a portion of a tensor at transcript concept (at a token ID).
Furthermore, THOMPSON TV does not disclose combining distinct transcript concepts across a conversation transcript to form concept combinations for populating structured data fields regardless of their locations in the transcript, nor does THOMPSON IV disclose concatenating tensor slices of those specific concept combinations. Paragraph [0074] of THOMPSON IV recites: "The individual token and snippet representations may include vectors and are sometimes referred to as embeddings. The cumulation or concatenation of these vectors or embeddings constitutes a tensor. The snippets and tokens may be referred to as tensors, because the snippets and/or tokens are typically batched and concatenated during training;". This batching of raw text tokens and snippets during offline training is structurally and functionally distinct from the claimed feature of concatenating specific tensor slices of a concept combination to form a concatenated tensor representation that is fed as an input to a machine learning model to evaluate validity. Because the applied references fail to disclose these structural elements, claim 1 and claim 10, and their corresponding dependent claims, are patentable over the applied references. Applicant respectfully requests withdrawal of the rejection under 35 U.S.C. § 103.
From Applicant’s argument above…
>>”Furthermore, THOMPSON TV does not disclose combining distinct transcript concepts across a conversation transcript to form concept combinations for populating structured data fields regardless of their locations in the transcript, nor does THOMPSON IV disclose concatenating tensor slices of those specific concept combinations.”<<
Distinct transcript concepts are not claimed and populating data fields regardless of their locations is not in claims 1 or 10 (claim 19).
From Applicant’s Claim 10…
“…determining, via the at least one computer processor, a plurality of concept combinations by combining the plurality of transcript concepts regardless of locations of two or more transcript concepts in the transcript,…”
Therefore, claim 10 recites combining concepts regardless of locations is claimed.
From Thompson…
“In some embodiments, the first portion of the classifier further includes a multi-headed intra-attention mechanism that aggregates, for each respective episodic record in the sub-plurality of episodic records, the corresponding plurality of corresponding contextualized token tensors for each respective snippet in the plurality of corresponding snippets to output a corresponding contextualized snippet tensor, thereby forming a corresponding plurality of corresponding contextualized snippet tensors for the respective episodic record.” [0031]
Therefore, a plurality of corresponding contextualized snippet tensors of an episodic record.
“An interaction between a patient and a healthcare provider is logged in a patient record, for example an electronic health record (EHR) or a hand-written record which may be later digitized to generate an electronic medical record (EMR). EHRs and EMRs are then stored in electronic medical system curated for the healthcare provider. These EHRs and EMRs typically have structured data, including medical codes used by the healthcare provider for billing purposes, and unrestructured data, including clinical notes and observations made by physicians, physician assistants, nurses, and others while attending to the patient.”
In further review of the claims and the prior art, and further search and consideration, while aspects of the invention are taught, the prior art rejection is withdrawn.
New Claims
New claims 19-24 depend from claim 1. Therefore, claims 19-24 are patentable for at least the reasons set forth above with respect to claim 1, and for the additional novel and nonobvious combinations recited therein.
In further review of the claims and the prior art, and further search and consideration, while aspects of the invention are taught, the prior art rejection is withdrawn
Claim Interpretation
Claims 1 and 10 recite the phrases of “via a machine learning model” and “via at least one computer processor.” For examination purposes, the word “via” is interpreted as “by.” Therefore, the phrases are interpreted as “by a machine learning model” and “by at least one computer processor.”
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, 4-10, and 13-24 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claims 1, 4-10, and 13-24 are directed to a system or method, which are statutory categories of invention. (Step 1: YES).
The Examiner has identified method Claim 10 as the claim that represents the claimed invention for analysis and is similar to system Claim 1.
Claim 10 recites the limitations of:
A method for determining structured data, the method comprising:
receiving, via at least one computer processor, a plurality of medical concepts and a corresponding plurality of labels, wherein each medical concept from the plurality of medical concepts is associated with a corresponding label from the plurality of labels;
receiving, via the at least one computer processor, a plurality of target attributes associated with a plurality of fields of one or more structured data, wherein each field from the plurality of fields is associated with one or more target attributes from the plurality of target attributes;
receiving, via the at least one computer processor, a transcript of a conversation between a physician and a patient;
determining, via the at least one computer processor, a plurality of transcript concepts in the transcript based on the plurality of medical concepts, where in each transcript concept from the plurality of transcript concepts is a corresponding medical concept from the plurality of medical concepts;
assigning, via the at least one computer processor, each transcript concept with the label associated with the corresponding medical concept;
determining, via the at least one computer processor, a plurality of concept combinations by combining the plurality of transcript concepts regardless of locations of two or more transcript concepts in the transcript, wherein each concept combination from the plurality of concept combinations is a combination of two or more transcript concepts from the plurality of transcript concepts, wherein the labels of the two or more transcript concepts associate with one or more target attributes from the plurality of target attributes of a same field from the plurality of fields;
for each concept combination of the plurality of concept combinations:
generating, via the at least one computer processor and using a pretrained language model, a tensor representation of at least a portion of the transcript that includes the two or more transcript concepts of the concept combination, wherein the tensor representation is stored in a computer memory and has dimensions of token by dimes, wherein the dims represent a length of word vectors of the pretrained language model,
determining, via the at least one computer processor, a plurality of slices of the tensor representation, wherein each slice of the plurality of slices corresponds to a token position of a respective one of the two or more transcript concepts of the concept combination to obtain a vector of length dims containing semantic and contextual properties of the respective transcript concept, and
concatenating, via the at least one computer processor, the plurality of slices to form a concatenated tensor representing the concept combination;
determining, via a machine learning model that received the concatenated tensor as an input, a combination label for each concept combination, the combination label being a valid label or an invalid label indicating whether the concept combination forms a valid instance of output structured data for the same field; and
generating, via the at least one computer processor, the output structured data based on the concept combinations having the valid label, wherein the output structured data is generated before the physician completes charting on the patient.
These above limitations, under their broadest reasonable interpretation, cover performance of the limitation as mental processes. The claim recites elements, in non-bold above, which covers performance of the limitation that can be concepts performed in the mind of a person or with pen and paper. A person can receive (read and comprehend mentally) medical concepts with labels, receive (read and comprehend) target attributes associated with fields, receive (read and comprehend) a transcript of a conversation between a physician and a patient. A person can determine (analyze in their mind by reading) transcript concepts in a transcript based on medical concepts, assign (mentally and with pen and paper) each transcript concept with a label corresponding to the medical concept, determine (analyze in their mind) a plurality of concept combinations by combining a plurality of concepts from a plurality of transcript concepts, generate (with pen and paper) a tensor representation of a portion of a transcript, determine (mentally and with pen and paper) slices of tensor representation to obtain vector length dims containing semantic and contextual properties, concatenate (with pen and paper) slices to form a tensor representing a combination, determine (with pen and paper) slices, determine (mentally) that the concatenated tensor forms a concept combination that is a valid or invalid label, generate an output (using pen and paper) of the concept combination having a valid label. If a claim limitation, under its broadest reasonable interpretation, covers performance of the limitation as a mental process, then it falls within the “Mental Processes” grouping of abstract ideas. Also, see MPEP 2106.04(a)(2) III C where using a generic computer was used in mental processes. Accordingly, the claim recites an abstract idea. Claim 1 is also abstract for similar reasons. (Step 2A-Prong 1: YES. The claims are abstract)
In that the claim recites a conversation between a physician and patient and provide an output to a user, the claim is also abstract as managing interactions or relationships between people, therefore abstract under Certain Methods of Organizing Human Activity.
Claim 10 recites the limitations of:
generating, via the at least one computer processor and using a pretrained language model, a tensor representation of at least a portion of the transcript that includes the two or more transcript concepts of the concept combination, wherein the tensor representation is stored in a computer memory and has dimensions of token by dimes, wherein the dims represent a length of word vectors of the pretrained language model,
determining, via the at least one computer processor, a plurality of slices of the tensor representation, wherein each slice of the plurality of slices corresponds to a token position of a respective one of the two or more transcript concepts of the concept combination to obtain a vector of length dims containing semantic and contextual properties of the respective transcript concept, and
concatenating, via the at least one computer processor, the plurality of slices to form a concatenated tensor representing the concept combination;
determining, via a machine learning model that received the concatenated tensor as an input, a combination label for each concept combination, the combination label being a valid label or an invalid label indicating whether the concept combination forms a valid instance of output structured data for the same field; and
The above limitations are also abstract as they are directed to mathematical concepts. Tensors are mathematical constructs representing or describing transcript concepts, in the form of vectors. Transforming text (transcript concepts) into mathematical concepts and processing (determining and concatenating) for input into a machine learning model is using a mathematical concept called tensors for processing text (see Applicant’s Remarks pg. 17 dated 07/07/2026).
This judicial exception is not integrated into a practical application. In particular, the claims only recite: computer-readable storage medium, computer processor, machine learning model, computer memory (Claim 1); computer processor, machine learning model, computer memory (Claim 10). The computer hardware is recited at a high-level of generality (i.e., as a generic processor performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. Also, the machine learning model is recited at a high-level of generality. See Applicant’s specification pg. 5, lines 10-17, where machine learning can be various models including decision tree and regression model, which can be performed with pen and paper (a person can create and analyze a decision tree and perform regression analysis). Accordingly, these additional elements, when considered separately and as an ordered combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. Therefore claims 1 and 10 are directed to an abstract idea without a practical application. (Step 2A-Prong 2: NO. The additional claimed elements are not integrated into a practical application)
The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception because, when considered separately and as an ordered combination, they do not add significantly more (also known as an “inventive concept”) to the exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of using a computer hardware amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. Accordingly, these additional elements, when considered separately and as an ordered combination, do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. Steps such as receiving and output (transmitting) are steps that are considered insignificant extra solution activity and mere instructions to apply the exception using general computer components (see MPEP 2106.05(d), II). Thus claims 1 and 10 are not patent eligible. (Step 2B: NO. The claims do not provide significantly more)
Dependent claims 4-9, and 13-24 further define the abstract idea that is present in their respective independent claims 1 and 10 and thus correspond to Mental Processes and Certain Methods of Organizing Human Activity and hence are abstract for the reasons presented above. The dependent claims do not include any additional elements that integrate the abstract idea into a practical application or are sufficient to amount to significantly more than the judicial exception when considered both individually and as an ordered combination. Claims 4 and 14 limit machine learning to a binary classifier, which is using existing classifier models at a high level of generality. Claims 5 and 14 recites using training data by a machine learning model at a high level of generality. Claims 6 and 15 recite annotate the transcript which is abstract as a mental process, but also recited at a high level of generality. Claims 7-9 and 16-17 are also abstract as mental processes, further limit abstract concepts and also recite output or highlight data at a high level of generality. Therefore, the claims 4-9 and 13-24 are directed to an abstract idea. Thus, the claims 1, 4-10, and 13-24 are not patent-eligible.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1, 4-10, and 13-24 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim 10 recites “determining, via the at least one computer processor, a plurality of slices of the tensor representation, wherein each slice of the plurality of slices corresponds to a token position of a respective one of the two or more transcript concepts of the concept combination to obtain a vector of length dims containing semantic and contextual properties of the respective transcript concept,…” where no teaching of “token position” can be found in the specification.
From Applicant’s only use of “token” in their specification…
“As used herein, the term "tensor" generally refers to a linear geometrical quantity. In the present description, the tensor is typically in the dimension of "[token, dims]", where "dims" is a length of word vectors that the text encoder (e.g., the text encoder 162) in use employs. Creating a slice means taking a portion of the tensor at a token ID corresponding to the transcript concept and obtaining the [dims] vector. This contains semantic and contextual information about a specific mention of the transcript concept.” (pg. 14, lines 19-24)
Therefore, a token ID corresponds to a transcript concept, not a token position. The “token position” is not taught in the disclosure.
A slice is created by taking a portion of a tensor at transcript concept (at a token ID). Claim 1 has a similar problem.
Claim 10 recites “generating, via the at least one computer processor, the output structured data based on the concept combinations having the valid label, wherein the output structured data is generated before the physician completes charting on the patient.
Claim 19 recites “… to render, concurrently with the plurality of fields, the transcript with the transcript concepts of each concept combination having the valid label visually highlighted and graphically linked to a respective field of the plurality of fields populated by the concept combination, wherein the plurality of fields are presented to the physician for review before the physician completes charting on the patient.
Claim 20 recites “… further configured to filter the transcript by discarding portions of the transcript that do not contain any of the plurality of transcript concepts prior to generating the tensor representation, thereby reducing a computational load of the pretrained language model and the machine learning model relative to processing an entirety of the transcript, and wherein the pretrained language model and the machine learning model are executed on the at least one computer processor with sufficient throughput to generate the output structured data before the physician completes charting on the patient.
From Applicant’s specification on charting…
“If, for example, this work was completed by hand and the aggregation of medical concepts was done via the same techniques disclosed herein, it could take potentially months to return the concepts to the physician, therefore rendering the concepts largely useless as they would no longer be actionable due to a change in the patient's status, medical condition, and the like. While concepts such as billing codes, can have more delay between the end of the encounter and when the computed result is returned, clinical concepts are largely viewed exclusively by the doctor, which can help achieve improved medical outcomes. This means that they need to be completed before the doctor completes charting on a given patient which often occurs during the course of the immediate or subsequent workday after the doctor patient/encounter. In contrast, given the large volume of data needed to utilize the techniques disclosed herein, it would take a human many weeks or months to generate the embedding for each identified concept, let alone statistically analyze each embedding pair making this approach without a machine learning model impracticable to perform by a human with or without the aid of pen and paper.” (pg. 21, lines 22-31 to pg. 22, lines 1-3)
Therefore, the aggregation of medical concepts needs to be completed before the doctor completes charting of the patient. There is no teaching that: 1) output structured data is generated before charting is completed; 2) the plurality of fields are presented to the physician for review before charting is completed; and 3) the language and machine learning model are executed with sufficient throughput to generate the output structured data before the physician completes charting. Claim 1 has a similar problem.
Claim 19 recites “configured to populate... and to render, concurrently with the plurality of fields, the transcript…” where no teaching of render concurrently with fields can be found in the specification.
Claim 19 recites “… the plurality of fields of the output structured data with the concept combinations having the valid label, and to render, concurrently with the plurality of fields, the transcript with the transcript concepts of each concept combination having the valid label visually highlighted and graphically linked to a respective field of the plurality of fields populated by the concept combination, wherein the plurality of fields are presented to the physician for review before the physician completes charting on the patient” where no teaching of label highlighted or linked is taught.
From Applicant’s specification…
“At step 216, the method 200 further includes generating, via the at least one computer processor 104, the output structured data 170 based on the concept combinations 142 having the valid label 154. In some embodiments, the method 200 further includes outputting the output structured data 170 to at least one of the user interface 172 and the at least one non-transitory computer-readable storage medium 102. In some embodiments, the method 200 further includes linking the output structured data 170 with the plurality of transcript concepts 130 in the transcript T. In some embodiments, the method 200 further includes visually highlighting the plurality of transcript concepts 130 in the transcript T linked to the output structured data 170.” (pg. 18, lines 29-31 to pg. 19, lines 2-5)
Therefore, transcripts concepts are highlighted (not labels) and the transcript is linked to the structured data (not label graphically linked to respective field of a plurality of fields).
Claim 20 recites “The system of claim 1, wherein the at least one computer processor is further configured to filter the transcript by discarding portions of the transcript that do not contain any of the plurality of transcript concepts prior to generating the tensor representation, thereby reducing a computational load of the pretrained language model and the machine learning model relative to processing an entirety of the transcript, and wherein the pretrained language model and the machine learning model are executed on the at least one computer processor with sufficient throughput to generate the output structured data before the physician completes charting on the patient.
Regarding filtering, the specification teaches filtering text of a transcript not filtering transcript and does not teach discarding portions of the transcript (pg. 8, lines 2-4 and pg. 19, lines 10-16). The specification also does not teach filtering prior to generating the tensor representation, thereby reducing computational load.
Regarding sufficient throughput, the specification does not teach sufficient throughput.
Claims 4-9 and 13-24 are further rejected as they depend from their respective independent claim.
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.
Claim 20 is 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.
Claim 20 recites “…and wherein the pretrained language model and the machine learning model are executed on the at least one computer processor with sufficient throughput to generate the output structured data before the physician completes charting on the patient” where sufficient is a relative term rendering the claim indefinite. Sufficient could be any amount of throughput.
Examiner Request
The Applicant is requested to indicate where in the specification there is support for amendments to claims should Applicant amend. The purpose of this is to reduce potential 35 U.S.C. §112(a) or §112 1st paragraph issues that can arise when claims are amended without support in the specification. The Examiner thanks the Applicant in advance.
Prior Art Analysis
A prior art search was conducted but does not result in a prior art rejection at this time. The best prior art found to date is Pub. No. US 2020/0185102 to Leventhal et al. combined with Pub. No. US 2024/0145050 to Thompson IV et al. While the combined references teach tensor, snippet (slices), machine learning, they do not teach all of the claim such as combining concepts regardless of location and output data before physician completes charting on patient.
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.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KENNETH BARTLEY whose telephone number is (571)272-5230. The examiner can normally be reached Mon-Fri: 7:30 - 4:00 EST.
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, SHAHID MERCHANT can be reached at (571) 270-1360. 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.
/KENNETH BARTLEY/Primary Examiner, Art Unit 3684
Prosecution Insights