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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 01/27/2026 was filed after the mailing date of the Non-Final Rejection on 01/07/2026. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Terminal Disclaimer
The terminal disclaimer filed on 04/07/2026 disclaiming the terminal portion of any patent granted on this application which would extend beyond the expiration date of any patent granted on Application 18/744,865 has been reviewed and is accepted. The terminal disclaimer has been recorded.
Response to Arguments
1. Regarding the objection to the drawings, Applicant has amended Fig. 2 to differentiate the processor and memory with different reference characters. Accordingly, the objection is withdrawn.
2. Regarding the rejection of claims 1-20 under 35 USC § 101, Applicant's arguments filed 04/07/2026 have been fully considered but they are not persuasive.
Applicant argues that on pgs. 8-9 of the Remarks that amended limitation of claim 1 of “generating, using the generative language model, a sequence of symbols identifying the one or more categories, the generation of the sequence based on the input and constrained by the grammar” cannot be performed in the human mind. The Examiner respectfully disagrees. Under Step 2A Prong 1, the step of “generating…a sequence of symbols identifying the one or more categories, the generation of the sequence based on the input and constrained by the grammar” can be performed by a person mentally with the aid of pen and paper. Specifically, a person can write down a sequence of symbols conforming to a grammar (particular set of rules/structure) and relating to the classification of the input. For example, a person can classify an input (e.g. a particular item is clothing), can obtain a grammar relating to this type (e.g. for items, using a format of item={category}), and then can write down an output identifying the category using the specified grammar/following the specific set of rules to write it (e.g. can write down specifically item = shoe). Therefore, the claim recites mental processes under Step 2A Prong 1. Under Step 2A Prong 2, additional elements are considered and viewed in combination with the claim as a whole to determine if the claims integrate the judicial exception into a practical application. The only additional limitation in the claim is the use of “a generative language model” to generate the sequence of symbols. This limitation is recited at a high level of generality and amounts to mere instructions to implement the judicial exception using a generic computer component. When viewed in combination with the claims as a whole, this limitation does not integrate the judicial exception into a practical application under Step 2A Prong 2 as it does not impose any meaningful limits on practicing the mental process. No technical details or specific model components are recited in the claim, and thus the claim is merely using a generic computer model to perform a process (observing, making a determination, and writing down information) which can be performed mentally, and thus does not reflect a technical improvement as argued.
Hence, Applicant’s arguments are not persuasive.
3. Regarding the provisional nonstatutory double patenting rejections, Applicant has filed a terminal disclaimer with the Remarks. Accordingly, the rejections are withdrawn.
4. Regarding the rejection of claims 1-2, 5-8, 10-13, and 15-20 under 35 U.S.C. § 102, and the rejections under 35 U.S.C. § 103, Applicant's arguments filed 04/07/2026 have been fully considered but they are not persuasive.
Applicant argues that amended claims 1, 12, and 20 are not anticipated by the prior art. Specifically, Applicant argues on pg. 10 that the amended limitation of “the generation of the sequence based on the input and constrained by the grammar” is not disclosed by the Scherle reference. The Examiner respectfully disagrees with this argument. Under the broadest reasonable interpretation of the claim, this limitation requires that an output be generated from a generative language model based on some input (e.g. prompt) and whose output is constrained (e.g. formatted) in a particular way. Scherle discloses this concept. Specifically, Scherle discloses an input to a generative language model specifying that the model generate output in a particular format, and automatically generating an output that is constrained to this format (para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure. …The data structure may, for example, be a JSON object structure.”; para. 0060 “The classification scheme (or another part of the instruction in the prompt data) may also specify a desired output format or structure for responses from the LLM 118.”). An LLM generating a response in a desired output format based on instructions in a prompt data reads on the BRI of “the generation of the sequence based on the input and constrained by the grammar”. Furthermore, the portion of the Scherle reference cited (para. 0062 cited on pg. 10 of Remarks) does not contradict this interpretation. Paragraph 0062 of the Scherle reference discloses a “postprocessing component 210 [which] postprocesses “raw” classification outputs from the LLM 118.” This postprocessing component checks/validates a model output to ensure that the output satisfies the grammar (para. 0062 “For example, the postprocessing component 210 may validate model output by checking whether the output has an expected structure (e.g., JSON object structure)…”). In other words, the postprocessing component is not what is generating the constrained sequence of symbols (this would instead be LLM 118). Instead, the postprocessing component is a secondary check verifying that an already constrained sequence of symbols (the response from LLM 118) is correct. Therefore, Scherle discloses constrained generation of a sequence of symbols using a generative language model, input, and grammar, as recited in amended claim 1.
Hence, Applicant’s arguments are not persuasive.
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.
5. Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1, “A computer-implemented method” is recited, which is directed to one of the four statutory categories of invention (process) (Step 1: YES). However, the claims limitations, under their broadest reasonable interpretation, recite mental processes which fall into the category of abstract idea (Step 2A Prong 1: YES).
The following limitations, under their broadest reasonable interpretation, recite mental processes:
receiving a prompt that instructs a generative language model to classify an input to the generative language model: a person writes down a prompt that can be input to a generative LM
obtaining a grammar responsive to the prompt, the grammar defining valid sequences of symbols corresponding to a plurality of categories, wherein the input can be classified into one or more of the plurality of categories: a person reads the prompt, and finds a grammar to use (e.g. set of rules for the structure of an output), where the output relates to a category of an input (e.g. type of item)
generating … a sequence of symbols identifying the one or more categories, the generation of the sequence based on the input and constrained by the grammar: a person writes down a sequence of symbols conforming to the grammar (structure) and relating to the classification of the input into a category (e.g. if the grammar is in form of “item = {category}”, a person observes input and writes down symbols which correspond to text, for example “item = shoe”)
Claim 1 does not contain any additional elements which integrate the judicial exception into a practical application (Step 2A Prong 2: NO). The only additional limitation is “generating, using the generative language model”. This limitation is recited at a high level of generality and amounts to mere instructions to implement the judicial exception using a generic computer, which do not integrate the judicial exception into a practical application as they do not impose any meaningful limits on practicing the abstract idea. Therefore, claim 1 is directed to an abstract idea.
Claim 1 does not contain any additional elements which amount to significantly more than the judicial exception (Step 2B: NO). The only additional limitations as discussed above amounts to mere instructions to implement the judicial exception using a generic computer, which does not amount to significantly more than the judicial exception as it does not provide an inventive concept. Therefore, claim 1 is not patent eligible.
Regarding claims 2-11, “The computer-implemented method” is recited, which is directed to one of the four statutory categories of invention (process) (Step 1: YES). However, the claims limitations, under their broadest reasonable interpretation, recite further mental processes or mathematical concepts which fall into the category of abstract idea (Step 2A Prong 1: YES).
The following limitations, under their broadest reasonable interpretation, recite mental processes or mathematical concepts:
Claim 2:
outputting an indication of the one or more categories into which the input has been classified based on the sequence of symbols: a person writes down the categories the input has been classified into (e.g. input = clothing, shoe, mens)
No additional limitations.
Claim 3:
wherein the symbols comprise tokens, and wherein generating the sequence of symbols includes: generating a plurality of values using the generative language model, each of the values indicative of a probability of a respective token being a next token of the sequence; applying a mask to the plurality of values, the mask operating on each value that corresponds to a token not compliant with the grammar to reduce or zero the probability of the token being the next token; and determining the next token based on the plurality of values after the mask is applied: generating a plurality of values of next token probabilities, applying a mask to reduce or zero probabilities of tokens not compliant, and to determine next token based on the application of the mask amounts to a mathematical concept.
No additional limitations.
Claim 4:
wherein the sequence of symbols, when mapped to text, provides a written indication of the one or more categories: a person writes the symbols (e.g. numbers) which map to text and express the category (e.g. ‘120’ corresponds to ‘men’s clothing’ category)
No additional limitations.
Claim 5:
wherein the prompt further comprises an instruction, and wherein obtaining the grammar responsive to the prompt further comprises obtaining the grammar based on the instruction: a person writes a prompt including an instruction, and obtains the grammar based on the instruction (e.g. person writes a prompt “What kind of clothing item is this?”, and selects a grammar related to apparel)
No additional limitations.
Claim 6:
wherein the instruction comprises information associated with the plurality of categories; and wherein obtaining the grammar based on the instruction further comprises: encoding the information associated with the plurality of categories within the grammar, wherein the encoding is in a format for parsing: and person writes down information about the categories within the grammar which can later be parsed (read and interpreted)
No additional limitations.
Claim 7:
wherein the grammar further comprises a label representative of the plurality of categories to which the grammar relates; and wherein obtaining the grammar based on the instruction further comprises: determining that the instruction is associated with the label; and selecting the grammar from a set of one or more grammars: a person determines the instruction relates to a label for a plurality of categories (e.g. label = clothing), and in response selects a grammar relating to ‘clothing’
No additional limitations.
Claim 8:
… comprises information associated with the plurality of categories; and wherein obtaining the grammar based on the instruction further comprises: retrieving the information associated with the plurality of categories …; and encoding the information associated with the plurality of categories within the grammar, wherein the encoding is in a format for parsing: a person obtains the grammar by obtaining information of the plurality of categories, and encodes the information in the grammar in a form which can be parsed
Claim 8 contains the additional limitation “a memory” and “retrieving…from the memory”, which amounts to mere instructions to implement the judicial exception using a generic computer.
Claim 9:
receiving an update to the plurality of categories, the update including at least one of an addition of a new category to the plurality of categories, a removal of a category from the plurality categories, or a modification of a category within the plurality of categories; and modifying the valid sequences of symbols in the grammar based on the update to the plurality of categories: a person adds, removes, or modifies a category, and reflect the change in the grammar (for example, adding a new valid category ‘shirt’ to a clothing grammar)
No additional limitations.
Claim 10:
Claim 10 contains the additional limitation “wherein the generative language model is a large language model”, which amounts to mere instructions to implement the judicial exception using generic computer components.
Claim 11:
wherein the grammar further constrains the valid sequences of symbols to a syntax of a programming language; and wherein outputting an indication of the one or more categories into which the input has been classified based on the sequence of symbols comprises outputting code of the programming language: a person writes down a classification which conforms to a programming language syntax (for example, writes down a JSON format reflecting the categories)
No additional limitations
Claims 2-11 do not contain any additional elements which integrate the judicial exception into a practical application (Step 2A Prong 2: NO). The only additional limitations amount to mere instructions to implement the judicial exception using a generic computer, which even when viewed in combination do not integrate the judicial exception into a practical application as they do not impose any meaningful limits on practicing the abstract idea. Therefore, claims 2-11 are directed to an abstract idea.
Claims 2-11 do not contain any additional elements which amount to significantly more than the judicial exception (Step 2B: NO). The only additional limitations as discussed above amount to mere instructions to implement the judicial exception using a generic computer, which even when viewed in combination do not amount to significantly more than the judicial exception as it does not provide an inventive concept. Therefore, claims 2-11 are not patent eligible.
Regarding claim 12, “A system” is recited, which is directed to one of the four statutory categories of invention (machine) (Step 1: YES). However, the claims limitations, under their broadest reasonable interpretation, recite limitations similar to those recited in method claim 1, and thus also mental processes which fall into the category of abstract idea (Step 2A Prong 1: YES) (see claim 1 analysis).
Claim 12 does not contain any additional elements which integrate the judicial exception into a practical application (Step 2A Prong 2: NO). The only additional limitations are “a memory to store a grammar”, “at least one processor to” and “generate, using the generative language model”. These limitations are recited at a high level of generality and amounts to mere instructions to implement the judicial exception using a generic computer, which do not integrate the judicial exception into a practical application as they do not impose any meaningful limits on practicing the abstract idea. Therefore, claim 12 is directed to an abstract idea.
Claim 12 does not contain any additional elements which amount to significantly more than the judicial exception (Step 2B: NO). The only additional limitations as discussed above amount to mere instructions to implement the judicial exception using a generic computer, which does not amount to significantly more than the judicial exception as it does not provide an inventive concept. Therefore, claim 12 is not patent eligible.
Regarding claims 13-19, “The system” is recited, which is directed to one of the four statutory categories of invention (machine) (Step 1: YES). However, the claims limitations, under their broadest reasonable interpretation, recite limitations similar to those recited in dependent claims 2, 3, 5-8, and 10, and thus also recite further mental processes or mathematical concepts which fall into the category of abstract idea (Step 2A Prong 1: YES) (see above analysis).
Claims 13-19 do not contain any additional elements which integrate the judicial exception into a practical application (Step 2A Prong 2: NO). The only additional limitations amount to mere instructions to implement the judicial exception using a generic computer, which even when viewed in combination do not integrate the judicial exception into a practical application as they do not impose any meaningful limits on practicing the abstract idea. Therefore, claims 13-19 are directed to an abstract idea.
Claims 13-19 do not contain any additional elements which amount to significantly more than the judicial exception (Step 2B: NO). The only additional limitations as discussed above amount to mere instructions to implement the judicial exception using a generic computer, which even when viewed in combination do not amount to significantly more than the judicial exception as it does not provide an inventive concept. Therefore, claims 13-19 are not patent eligible.
Regarding claim 20, “One or more non-transitory computer readable media” is recited, which is directed to one of the four statutory categories of invention (article of manufacture) (Step 1: YES). However, the claims limitations, under their broadest reasonable interpretation, recite limitations similar to those recited in method claim 1, and thus also mental processes which fall into the category of abstract idea (Step 2A Prong 1: YES) (see claim 1 analysis).
Claim 20 does not contain any additional elements which integrate the judicial exception into a practical application (Step 2A Prong 2: NO). The only additional limitations are “One or more non-transitory computer readable media having stored thereon computer-executable instructions that, when executed by at least one computer, cause the at least one computer to perform a method comprising” and “generating, using the generative language model”. These limitations are recited at a high level of generality and amounts to mere instructions to implement the judicial exception using a generic computer, which do not integrate the judicial exception into a practical application as they do not impose any meaningful limits on practicing the abstract idea. Therefore, claim 20 is directed to an abstract idea.
Claim 20 does not contain any additional elements which amount to significantly more than the judicial exception (Step 2B: NO). The only additional limitations as discussed above amount to mere instructions to implement the judicial exception using a generic computer, which does not amount to significantly more than the judicial exception as it does not provide an inventive concept. Therefore, claim 20 is not patent eligible.
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
6. Claims 1-2, 5-8, 10-13, and 15-20 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Scherle (US 2025/0173550 A1).
Regarding claim 1, Scherle discloses A computer-implemented method (Abstract) comprising: receiving a prompt that instructs a generative language model to classify an input to the generative language model (para. 0053 “The LLM 118 may provide natural language processing capabilities that enable it to respond to various queries. Specifically, the LLM 118 may process a prompt provided by the data classification system 128 and automatically classify attributes included in the prompt.”); obtaining a grammar responsive to the prompt (para. 0031 “The request included in the prompt data may include a predetermined message (e.g., a static system message) that instructs the machine learning model with respect to its task.”; para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure.”), the grammar defining valid sequences of symbols corresponding to a plurality of categories, wherein the input can be classified into one or more of the plurality of categories (para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure.”; para. 0022 “The classification scheme may be used to guide or indicate to a machine learning model how to classify the process metadata. For example, the classification scheme may indicate that each attribute should be classified as either being a personal data attribute or not being a personal data attribute. Alternatively or additionally, the classification scheme may indicate that each attribute should be classified into one of a set of predetermined categories (e.g., “text,” “numerical,” “categorical,” or “date”). The classification scheme may provide a standardized taxonomy to categorize items using consistent labels across different automated processes (e.g., within an organization). The instruction may thus include or reference the predetermined classification scheme.”); and generating, using the generative language model, a sequence of symbols identifying the one or more categories (para. 0023 “The method may include providing or transmitting the prompt data to the machine learning model to obtain output comprising, for each of the plurality of attributes, a classification result. The classification result may include a personal data indicator that indicates whether the attribute is a personal data attribute. The classification result may include a category of the attribute as identified by the machine learning model (e.g., from among a plurality of data candidate categories included in the prompt data).”), the generation of the sequence based on the input (para. 0060 “In some examples, the prompt generating component 206 adds an instruction to classify the process metadata according to a predetermined classification scheme.”; para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure.”; para. 0060 “The classification scheme (or another part of the instruction in the prompt data) may also specify a desired output format or structure for responses from the LLM 118.”) and constrained by the grammar (para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure.”).
Regarding claim 2, Scherle discloses further comprising outputting an indication of the one or more categories into which the input has been classified based on the sequence of symbols (para. 0023 “The method may include providing or transmitting the prompt data to the machine learning model to obtain output comprising, for each of the plurality of attributes, a classification result. The classification result may include a personal data indicator that indicates whether the attribute is a personal data attribute. The classification result may include a category of the attribute as identified by the machine learning model (e.g., from among a plurality of data candidate categories included in the prompt data).”).
Regarding claim 5, Scherle discloses wherein the prompt further comprises an instruction (para. 0031 “The request included in the prompt data may include a predetermined message (e.g., a static system message) that instructs the machine learning model with respect to its task.”), and wherein obtaining the grammar responsive to the prompt further comprises obtaining the grammar based on the instruction (para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure.”; para. 0060 “In some examples, the prompt generating component 206 adds an instruction to classify the process metadata according to a predetermined classification scheme.”).
Regarding claim 6, Scherle discloses wherein the instruction comprises information associated with the plurality of categories (para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure.”; para. 0060 “In some examples, the prompt generating component 206 adds an instruction to classify the process metadata according to a predetermined classification scheme.”; para. 0022 “Alternatively or additionally, the classification scheme may indicate that each attribute should be classified into one of a set of predetermined categories (e.g., “text,” “numerical,” “categorical,” or “date”). The classification scheme may provide a standardized taxonomy to categorize items using consistent labels across different automated processes (e.g., within an organization). The instruction may thus include or reference the predetermined classification scheme.”); and wherein obtaining the grammar based on the instruction further comprises: encoding the information associated with the plurality of categories within the grammar, wherein the encoding is in a format for parsing (information associated with categories encoded within the grammar (JSON format): para. 0068-0078 “The process design application 126 may store the attributes in the database 132. For example, a scheme file containing the attributes may include attribute names and attribute data types, as shown below for an exemplary loan application process. The example below is simplified, and it will be appreciated that the scheme file may include many more attributes, as well as other metadata. [0069] Loan classification [0070] { [0071] “name”: “STRING”, [0072] “lastName”: “STRING”, [0073] “age”: “INT”, [0074] “amount”: “DOUBLE”, [0075] “address”: “STRING”, [0076] “marriageState”: “STRING” [0077] . . . [0078] }”).
Regarding claim 7, Scherle discloses wherein the grammar further comprises a label representative of the plurality of categories to which the grammar relates (instruction includes reference to a particular scheme composed of category: para. 0022 “The classification scheme may provide a standardized taxonomy to categorize items using consistent labels across different automated processes (e.g., within an organization). The instruction may thus include or reference the predetermined classification scheme.”); and wherein obtaining the grammar based on the instruction further comprises: determining that the instruction is associated with the label (the reference is included in the instruction added to prompt: para. 0022 “The classification scheme may provide a standardized taxonomy to categorize items using consistent labels across different automated processes (e.g., within an organization). The instruction may thus include or reference the predetermined classification scheme.”); and selecting the grammar from a set of one or more grammars (uses specified classification scheme as the grammar for output: para. 0060 “The classification scheme may thus provide a common “vocabulary” for the machine learning model to use when classifying the attributes. The results are output using categories and labels as defined in the classification scheme.”).
Regarding claim 8, Scherle discloses wherein a memory comprises information associated with the plurality of categories (para. 0068 “The process design application 126 may store the attributes in the database 132. For example, a scheme file containing the attributes may include attribute names and attribute data types, as shown below for an exemplary loan application process. The example below is simplified, and it will be appreciated that the scheme file may include many more attributes, as well as other metadata. [0069] Loan classification [0070] { [0071] “name”: “STRING”, [0072] “lastName”: “STRING”, [0073] “age”: “INT”, [0074] “amount”: “DOUBLE”, [0075] “address”: “STRING”, [0076] “marriageState”: “STRING” [0077] . . . [0078] }”); and wherein obtaining the grammar based on the instruction further comprises: retrieving the information associated with the plurality of categories from the memory (para. 0068 “The process design application 126 may store the attributes in the database 132. For example, a scheme file containing the attributes may include attribute names and attribute data types, as shown below for an exemplary loan application process. The example below is simplified, and it will be appreciated that the scheme file may include many more attributes, as well as other metadata. [0069] Loan classification [0070] { [0071] “name”: “STRING”, [0072] “lastName”: “STRING”, [0073] “age”: “INT”, [0074] “amount”: “DOUBLE”, [0075] “address”: “STRING”, [0076] “marriageState”: “STRING” [0077] . . . [0078] }”; para. 0057 “The metadata access component 202 accesses or retrieves process metadata to be analyzed by the LLM 118. For example, the metadata access component 202 may access a scheme file that contains the process metadata (e.g., select a scheme file that is stored in association with an identifier of a particular automated process)...The metadata access component 202 may retrieve the process metadata from a storage location, such as the database 132 of FIG. 1.”); and encoding the information associated with the plurality of categories within the grammar, wherein the encoding is in a format for parsing (information associated with categories encoded in a grammar (JSON format): para. 0068 “The process design application 126 may store the attributes in the database 132. For example, a scheme file containing the attributes may include attribute names and attribute data types, as shown below for an exemplary loan application process. The example below is simplified, and it will be appreciated that the scheme file may include many more attributes, as well as other metadata. [0069] Loan classification [0070] { [0071] “name”: “STRING”, [0072] “lastName”: “STRING”, [0073] “age”: “INT”, [0074] “amount”: “DOUBLE”, [0075] “address”: “STRING”, [0076] “marriageState”: “STRING” [0077] . . . [0078] }”).
Regarding claim 10, Scherle discloses wherein the generative language model is a large language model (LLM) (para. 0053 “The LLM 118 may provide natural language processing capabilities that enable it to respond to various queries. Specifically, the LLM 118 may process a prompt provided by the data classification system 128 and automatically classify attributes included in the prompt.”).
Regarding claim 11, Scherle discloses wherein the grammar further constrains the valid sequences of symbols to a syntax of a programming language; and wherein outputting an indication of the one or more categories into which the input has been classified based on the sequence of symbols comprises outputting code of the programming language (Scherle discloses constraint of LLM output to syntax of JSON; JSON is being interpreted as reading on BRI of “a programming language” in light of Applicant’s specification (see para. 0109, 0115); Scherle, para. 0032 “Accordingly, the instruction included in the prompt data may identify a data structure for returning the classification results, with the output automatically being structured according to the identified data structure. The method may include providing the process design application with programmatic access to the output to integrate the classification results into the automated process (e.g., to allow the personal data indicators to be automatically applied when the automated process runs). The data structure may, for example, be a JSON object structure.”).
Regarding claim 12, claim 12 is a system claim with limitations similar to method claim 1, and is thus rejected under similar rationale.
Additionally, Scherle discloses A system (Fig. 8) comprising: a memory to store a grammar (Fig. 8, 804, para. 0059 “For example, the prompt generating component 206 may add a natural language instruction stored in the database 132 to the process metadata to provide a task, a role, constraints, or additional context to the LLM 118.”); and at least one processor to (Fig. 8, 802).
Regarding claim 13, claim 13 is rejected for analogous reasons to claim 2.
Regarding claim 15, claim 15 is rejected for analogous reasons to claim 5.
Regarding claim 16, claim 16 is rejected for analogous reasons to claim 6.
Regarding claim 17, claim 17 is rejected for analogous reasons to claim 7.
Regarding claim 18, claim 18 is rejected for analogous reasons to claim 8.
Regarding claim 19, claim 19 is rejected for analogous reasons to claim 10.
Regarding claim 20, claim 20 is a non-transitory computer media claim with limitations similar to method claim 1, and is thus rejected under similar rationale.
Additionally, Scherle discloses One or more non-transitory computer readable media having stored thereon computer-executable instructions that, when executed by at least one computer, cause the at least one computer to perform a method comprising (para. 0134 “Example 18 is a non-transitory computer-readable medium that stores instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising…”).
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.
7. Claims 3-4 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Scherle in view of Willard & Louf (NPL Efficient Guided Generation for Large Language Models, hereinafter Willard).
Regarding claim 3, Scherle discloses a generative language model but does not specifically disclose generating a plurality of values using the generative language model, each of the values indicative of a probability of a respective token being a next token of the sequence; applying a mask to the plurality of values, the mask operating on each value that corresponds to a token not compliant with the grammar to reduce or zero the probability of the token being the next token; and determining the next token based on the plurality of values after the mask is applied.
Willard teaches generating a plurality of values using the generative language model, each of the values indicative of a probability of a respective token being a next token of the sequence (pg. 2, section 2 “We define the next token st+1 as the following random variable: α = LLM (St, θ) st+1 ~ Categorical (α)”); applying a mask to the plurality of values, the mask operating on each value that corresponds to a token not compliant with the grammar to reduce or zero the probability of the token being the next token (pg. 4, section 2.2 “Since we are dealing with a finite, discrete distribution, we can compute an un-normalized conditional distribution by applying a Boolean mask m: P(V) → {0,1}N that restricts the support of the original distribution…”); and determining the next token based on the plurality of values after the mask is applied (pg. 4, section 2.2 “The resulting conditional distribution implied by st+1 encodes constraints on the support of s t+1. For instance, the masks m could be designed so that the generated sequences…represent digit samples, strings that match the regular expression [a-zA-Z], and strings that parse according to a specified grammar (e.g. Python, SQL, etc.)”).
Scherle and Willard are considered to be analogous to the claimed invention as
they both are in the same field of large language model generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Scherle to incorporate the teachings of Willard in order to generate a plurality of values indicative of a probability of a respective token being a next token, applying a mask to the plurality of values to a token not compliant with the grammar to reduce or zero the probabilities of the token being the next token, and to determine the next token based on the plurality of values after the mask is applied. Doing so would be beneficial, as this would enable LLMs to be used for tasks requiring rigid formatting requirements that are either hard or costly to capture through fine-tuning alone (pg. 1, section 1).
Regarding claim 4, Scherle discloses a step of provides a written indication of the one or more categories (para. 0023 “The method may include providing or transmitting the prompt data to the machine learning model to obtain output comprising, for each of the plurality of attributes, a classification result. The classification result may include a personal data indicator that indicates whether the attribute is a personal data attribute. The classification result may include a category of the attribute as identified by the machine learning model (e.g., from among a plurality of data candidate categories included in the prompt data).”). Scherle does not specifically disclose wherein the sequence of symbols, when mapped to text, [provides a written indication…].
Willard discloses a grammar applied to a sequence of symbols (pg. 2, section 2.2, applying Boolean mask to next-token distribution sequence) and wherein the sequence of symbols, when mapped to text, provides a written indication (pg. 2, section 2, sequence of symbols St corresponds to a vocabulary V composed of strings from a fixed alphabet).
Scherle and Willard are considered to be analogous to the claimed invention as
they both are in the same field of large language model generation. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Scherle to incorporate the teachings of Willard in order to have the sequence of symbols conforming to the grammar map to text to provide a written description. Doing so would be beneficial, as this would enable LLMs to be used for tasks requiring rigid formatting requirements that are either hard or costly to capture through fine-tuning alone (pg. 1, section 1).
Regarding claim 14, claim 14 is rejected for analogous reasons to claim 3.
8. Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Scherle in view of Arcadihno et al. (US 2023/0325154 A1, hereinafter Arcadihno).
Regarding claim 9, Scherle does not specifically disclose receiving an update to the plurality of categories, the update including at least one of an addition of a new category to the plurality of categories, a removal of a category from the plurality categories, or a modification of a category within the plurality of categories; and modifying the valid sequences of symbols in the grammar based on the update to the plurality of categories.
Arcadihno teaches receiving an update to the plurality of categories, the update including at least one of an addition of a new category to the plurality of categories, a removal of a category from the plurality categories, or a modification of a category within the plurality of categories (para. 0048 “For a given query and data model schema pair, the grammar shown previously can be augmented with one or more additional rules. For example, two additional rules specifying the valid tables and the valid columns can be added. The following production rules may be added: {table-name} |= user | account {column-name} |= user.id | user.name | user.birth-date | user.name | user.country | account.userID | account.country”); and modifying the valid sequences of symbols in the grammar based on the update to the plurality of categories (para. 0043 “The constrained decoder 220 is configured to determine valid tokens based on one or more tokens that have been generated by the generator. The constrained decoder limits eligible options for a token to include in the output in the computer language based on a specified grammar for the computer language.”).
Scherle and Arcadihno are considered to be analogous to the claimed invention as they both are in the same field of generating grammar-constrained outputs. Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Scherle to incorporate the teachings of Arcadihno in order to receive an update to the plurality of categories, and to modify the valid sequences of symbols in the grammar based on the update to the plurality of categories. Doing so would allow for a more flexible generation of outputs dependent on the queries received (para. 0048).
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 CODY DOUGLAS HUTCHESON whose telephone number is (703)756-1601. The examiner can normally be reached M-F 8:00AM-5:00PM 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, 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.
/CODY DOUGLAS HUTCHESON/Examiner, Art Unit 2659
/PIERRE LOUIS DESIR/Supervisory Patent Examiner, Art Unit 2659
Prosecution Insights