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 .
Claims 1-3, 5-10, 12-17, and 19-20 are presented for examination.
Response to Amendment
Applicant’s amendment has obviated most, but not all, of the specification, drawing, and claim objections given in the last Office action. To the extent that an objection or rejection appears in the previous Office Action(s) but not this Office Action, that objection or rejection is withdrawn. To the extent that it appears both in a previous Office Action(s) and this Office Action, the objection or rejection is maintained.
Claim Objections
Claims 3, 10, and 17 are objected to because of the following informalities: “the one or more entity rule” should be “the entity rule”.
Appropriate correction is required.
Claim Rejections - 35 USC § 101
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
Claims 1-3, 5-10, 12-17, and 19-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis of the claims will follow the 2019 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (“2019 PEG”).
Claim 1
Step 1: The claim recites a system comprising a memory and a processor; therefore, it is directed to the statutory category of machines.
Step 2A Prong 1: The claim recites, inter alia:
[A]nalyz[ing] a first set of data objects associated with a first set of data groups to derive a set of logic relationships between the first set of data objects: This limitation could encompass mentally analyzing the data objects and mentally deriving the logical relationships among them.
[D]etermin[ing] whether the machine learning model is trained to meet an entity rule of the plurality of entity rules in accordance with a desired accuracy: This limitation could encompass mentally determining whether the model is sufficiently trained by visually observing its outputs.
[D]etermin[ing] whether the tested machine learning model meets one or more regulation rules and one or more core values: This limitation could encompass mentally determining whether the model meets regulations and values by observing the model.
[I]n response to determining that the machine learning model satisfies the entity rule in accordance with the desired accuracy: … generat[ing] a threat scenario associated with the first set of data objects based at least on [a] test: This limitation could encompass mentally generating the threat scenarios.
[D]etermin[ing] a risk level of the threat scenario based at least in part on a risk matrix: This limitation could encompass mentally determining the risk levels based on a risk matrix.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “a memory operable to store: a plurality of data groups comprising a plurality of data objects; a plurality of entity rules; a plurality of regulation rules; and one or more core values; and a processor operably coupled to the memory”. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f). The claim further recites “train[ing], based on a set of training rules, a machine learning model with the first set of data objects and the set of logic relationships” and “test[ing], utilizing a generative network, the machine learning model based at least in part on the first set of data objects, the set of logic relationships, a regulation rule of the plurality of regulation rules, and a core value of the plurality of core values”. However, these limitations merely restrict the field of use of the judicial exception to model training and testing. MPEP § 2106.05(h). Finally, the claim’s recitation of “in response to determining that the tested machine learning model satisfies the regulation rule and the core value, deploy[ing] the tested machine learning model into a real-time application” amounts to the insignificant post-solution activity of mere data gathering and output. MPEP § 2106.05(g).
Step 2B: The claim does not contain significantly more than the judicial exception. The analysis at this step mirrors that of step 2A, prong 2, except insofar as the deploying limitation, in addition to reciting insignificant extra-solution activity, also recites the well-understood, routine, and conventional activity of receiving or transmitting data over a network. MPEP § 2106.05(d)(II); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network). As an ordered whole, the claim is directed to a mentally performable algorithm for analyzing a model to determine whether it meets a set of rules and values and generating additional data if it does not. Nothing in the claim provides significantly more than this. As such, the claim is not patent eligible.
Claim 2
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites, inter alia, “in response to determining that the tested machine learning model fails to satisfy the regulation rule and the core value: refin[ing] one or more training rules of the set of training rules to generate a second set of data objects associated with a second set of data groups”. This limitation could encompass mentally determining that the model does not meet the regulation rules or core values and mentally generating the second data objects by mentally refining the training rules.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that “the processor is configured” to perform the method. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f). The claim further recites “retrain[ing] the tested machine learning model based at least in part on the second set of the data objects to satisfy the regulation rule and the core value.” However, this limitation merely restricts the field of use of the judicial exception to model retraining. MPEP § 2106.05(h).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that “the processor is configured” to perform the method. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f). The claim further recites “retrain[ing] the tested machine learning model based at least in part on the second set of the data objects to satisfy the regulation rule and the core value.” However, this limitation merely restricts the field of use of the judicial exception to model retraining. MPEP § 2106.05(h).
Claim 3
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites, inter alia, “ in response to determining that the tested machine learning model fails to satisfy the entity rules with the desired accuracy: refine one or more training rules of the set of training rules to generate a third set of data objects associated with a third set of data groups”. This limitation could encompass mentally determining that the model does not meet the rules by observation of the model, then mentally refining the training rules and mentally generating the set of objects.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites “retrain[ing] the tested machine learning model with the third set of data objects to meet the one or more entity rule[s] with the desired accuracy.” However, this limitation merely restricts the field of use of the judicial exception to model retraining. MPEP § 2106.05(h).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites “retrain[ing] the tested machine learning model with the third set of data objects to meet the one or more entity rule[s] with the desired accuracy.” However, this limitation merely restricts the field of use of the judicial exception to model retraining. MPEP § 2106.05(h).
Claim 5
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites that “refin[ing] the one or more training rules by adjusting one or more hyperparameters, adjusting sampling rules, or collecting additional historical data.” This limitation could encompass mentally making adjustments to the sampling rules.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that “the processor is further configured” to perform the method. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that “the processor is further configured” to perform the method. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim 6
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites that “the sampling rules comprise stratified sampling, cluster sampling or random sampling.” The improvement of the sampling rules remains mentally performable under these further assumptions.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. See claim 5 analysis.
Step 2B: The claim does not contain significantly more than the judicial exception. See claim 5 analysis.
Claim 7
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites that “collect[ing] the additional historical data by collecting augmented data or rescaling the additional historical data associated with the plurality of data groups.” This limitation could encompass mentally rescaling the data.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites that “the processor is further configured” to perform the method. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The claim further recites that “the processor is further configured” to perform the method. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claims 8-10, 12-14
Step 1: The claims recite a method; therefore, they are directed to the statutory category of processes.
Step 2A Prong 1: The claims recite the same judicial exceptions as in claims 1-3 and 5-7, respectively.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The analysis at this step mirrors that of claims 1-3 and 5-7, respectively, except insofar as these claims do not recite the hardware recited therein.
Step 2B: The claim does not contain significantly more than the judicial exception. The analysis at this step mirrors that of claims 1-3 and 5-7, respectively, except insofar as these claims do not recite the hardware recited therein.
Claims 15-17, 19-20
Step 1: The claims recite a non-transitory computer-readable medium; therefore, the claims are directed to the statutory category of articles of manufacture.
Step 2A Prong 1: The claims recite the same judicial exceptions as in claims 1-3, 5, and 6-7 combined, respectively.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The analysis at this step mirrors that of claims 1-3, 5, and 6-7 combined, respectively, except insofar as these claims recite a “non-transitory computer-readable medium storing instructions that when executed by a processor causes the processor to [perform the method]”. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Step 2B: The claim does not contain significantly more than the judicial exception. The analysis at this step mirrors that of claims 1-3, 5, and 6-7 combined, respectively, except insofar as these claims recite a “non-transitory computer-readable medium storing instructions that when executed by a processor causes the processor to [perform the method]”. However, this is a mere instruction to apply the judicial exception using a generic computer. MPEP § 2106.05(f).
Claim Rejections - 35 USC § 103
Claims 1-3, 5, 7-10, 12, 14-17, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Wodetski et al. (US 20220398680) (“Wodetski”) in view of Sharma Mittal et al. (US 20230128548) (“Sharma”) and further in view of Hoque et al., “Risk-Ranking Matrix for Security Patching of Exploitable Vulnerabilities,” in 2808.1 Proc. 1st Int’l Conf. Frontier of Digital Tech. Towards a Sustainable Soc’y 050004 (2023) (“Hoque”) and Madani et al. (US 20190197368) (“Madani”).
Regarding claim 1, Wodetski discloses “[a] system comprising:
a memory (processor may include memory that stores methods, codes, instructions, and programs – Wodetski, paragraph 160) operable to store:
a plurality of data groups comprising a plurality of data objects (Wodetski Fig. 8 shows that sample contracts [data objects] are processed using ML/AI algorithms and pre-classified [analyzed/grouped] into contractual document types);
a plurality of entity rules (Wodetski Fig. 15 shows that machine learning training is applied until a performance/accuracy crosses a desired threshold, and high-performing universal contract model rules [entity rules]/models are packaged for deployment);
a plurality of regulation rules (industry-specific rules [core values] are optionally packaged to supplement universal rules, and customer-specific rules [regulation rules] are optionally packaged to supplement universal rules – Wodetski, paragraph 131); and
one or more core values (industry-specific rules [core values] are optionally packaged to supplement universal rules, and customer-specific rules [regulation rules] are optionally packaged to supplement universal rules – Wodetski, paragraph 131); and
a processor operably coupled to the memory (processor may include memory that stores methods, codes, instructions, and programs – Wodetski, paragraph 160) and configured to:
analyze a first set of data objects associated with a first set of data groups to derive a set of logic relationships between the first set of data objects (Wodetski Fig. 8 shows that sample contracts [data objects] are processed using ML/AI algorithms and pre-classified [analyzed/grouped] into contractual document types, especially contract transaction types (create contract, create order, amend, review, assign, terminate, etc.) [logic relationships between the data objects = classification in common]);
train, based on a set of training rules, a machine learning model with the first set of data objects and the set of logic relationships (Wodetski Fig. 8 shows that, once human experts review pre-classified documents, valid classifications are passed to the trained corpus [containing data objects/logic relationships] with classification annotations and the trained corpus is then used to train ML/AI models; Fig. 15 shows that training is applied to the corpus using multiple techniques [training rules] until performance/accuracy crosses a desired threshold);
determine whether the machine learning model is trained to meet an entity rule of the plurality of entity rules in accordance with a desired accuracy (Wodetski Fig. 15 shows that machine learning training is applied until a performance/accuracy crosses a desired threshold, and high-performing universal contract model rules [entity rules]/models are packaged for deployment); …
determining that the machine learning model meets one or more entity rules with the desired accuracy (Wodetski Fig. 15 shows that machine learning training is applied until a performance/accuracy crosses a desired threshold, and high-performing universal contract model rules [entity rules]/models are packaged for deployment) … [and using] the first set of the data objects and the set of the logic relationships (see Wodetski Figs. 8 and 15 and note that both are used to train the model); …
a regulation rule of the one or more regulation rules, and a core value of the plurality of core values (high-performing universal contract model rules/models are packaged for deployment; industry-specific rules [core values] are optionally packaged to supplement universal rules, and customer-specific rules [regulation rules] are optionally packaged to supplement universal rules; the industry-specific rules and customer rules are deployed to an AI engine [i.e., the AI system then handles/meets these rules] – Wodetski, paragraphs 131-32) …; [and] …
deploy[ing] the … machine learning model into a real-time application (Wodetski paragraph 110 discloses that the platform [including the model] is deployed using a normalized data model; paragraph 138 discloses that the data used with the model are updated in real time).”
Wodetski appears not to disclose explicitly the further limitations of the claim. However, Sharma discloses “in response to determining that the machine learning model meets one or more … rules with the desired accuracy:
test[ing] … the machine learning model based at least in part on the first set of data objects … [and] a … rule of the plurality of … rules (generating the training dataset is an iterative process where data are received from data sources, aggregated by a central server, and used to generate a training dataset which is used to train the model; the model is then tested against a validation dataset [first set] to identify an accuracy of the trained machine-learning model; aggregated data are then used to train a local machine learning model to generate new data; these new data are used to refine the training dataset which is then used to retrain the machine learning model until the process converges; convergence occurs when the change to the statistics is minimal or under a threshold [rule] – Sharma, paragraph 51) …; … [and]
in response to determining that the tested machine learning model satisfies the … rule …, deploy[ing] the tested machine learning model into a[n] … application (aggregated data are then used to train a local machine learning model to generate new data; these new data are used to refine the training dataset which is then used to retrain the machine learning model until the process converges; convergence occurs when the change to the statistics is minimal or under a threshold [i.e., when the rule is satisfied, the model is deployed] – Sharma, paragraph 51).”
Sharma and the instant application both relate to machine learning and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wodetski to test the model and retrain it with new training data until a criterion is met, as disclosed by Sharma, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the model is up-to-date and better geared toward fulfilling its intended purpose, thereby increasing the accuracy of the model on the task. See Sharma, paragraph 51.
Neither Wodetski nor Sharma appears to disclose explicitly the further limitations of the claim. However, Hoque discloses “generat[ing] a threat scenario associated with the first set of the data objects based at least on [a] test (Table 2 of Hoque shows a risk ranking matrix that assigns a Common Vulnerability Scoring System (CVSS) score and a ranking to various threat scenarios considering the gained access type, confidentiality impact, integrity impact, availability impact, access complexity, and authentication [collectively comprising a threat scenario]; first paragraph under “A Proof-of-Concept Tool to Demonstrate Risk Ranking Matrix” discloses that the models assessing these scenarios are tested with real vulnerability data);
determin[ing] a risk level of the threat scenario based at least in part on a risk matrix (Table 2 of Hoque shows a risk ranking matrix that assigns a Common Vulnerability Scoring System (CVSS) score and a ranking [risk level] to various threat scenarios considering the gained access type, confidentiality impact, integrity impact, availability impact, access complexity, and authentication); and …
testing the [system] based at least in part on the threat scenario and the risk level (Hoque Table 3 and section entitled “Application of the Proposed Matrix on Test Data” show that risk matrix containing the threat scenarios was tested to determine the number of exploitable vulnerabilities in a national vulnerability database based on the ranking) ….”
Hoque and the instant application both relate to data security and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Wodetski and Sharma to analyze a data threat environment using a risk matrix, as disclosed by Hoque, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would save critical assets and objectives by providing an improved vulnerability risk ranking framework. See Hoque, Introduction.
Neither Wodetski, Sharma, nor Hoque appears to disclose explicitly the further limitations of the claim. However, Madani discloses “test[ing], utilizing a generative network, the … model (generative adversarial network [generative network] based framework generates medical image data and trains a medical image classifier based on an expanded medical image dataset [data] – Madani, paragraph 22; GAN was trained on Dataset 1 and tested using Dataset 2 [so the generator of the GAN is tested using the discriminator and vice versa] – id. at paragraph 66) …; … [and]
test[ing] the machine learning model (GAN was trained on Dataset 1 and tested using Dataset 2 – Madani, paragraph 66) ….”
Madani and the instant application both relate to generative networks and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Wodetski, Sharma, and Hoque to employ a tested generative network, as disclosed by Madani, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would allow the network to generate data rather than merely classifying them, thereby enhancing its performance capabilities. See Madani, paragraph 2.
Claim 8 is a method claim corresponding to system claim 1 and is rejected for the same reasons as given in the rejection of that claim. Similarly, claim 15 is a non-transitory computer-readable medium claim corresponding to system claim 1 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 2, the rejection of claim 1 is incorporated. Wodetski further discloses “the regulation rule and the core value”, as shown in the rejection of claim 1.
Sharma further discloses that “the processor is further configured to:
in response to determining that the tested machine learning model fails to satisfy the … rule …:
refin[ing] one or more training rules of the set of training rules to generate a second set of data objects associated with a second set of data groups (aggregated data are then shared with data sources who use the aggregated data to train a local machine-learning model to generate new data to be provided to a central server; these new data are used to refine the training dataset [refined training set = second set of data objects] which is then used to retrain the machine-learning model; the machine-learning model is tested against the validation dataset and this process continues until the process converges [determining that the process has not converged = determining that the model does not meet the rule that the process should converge; the fact that particular new data should be added to the training dataset may be regarded as a training rule, so that rule is refined when specific new data are added] – Sharma, paragraph 51; training dataset includes labels [data groups] identifying a correct classification for the datapoint – id. at paragraph 1); and
retrain[ing] the tested machine learning model based at least in part on the second set of the data objects to satisfy the … rule (aggregated data are then shared with data sources who use the aggregated data to train a local machine-learning model to generate new data to be provided to a central server; these new data are used to refine the training dataset [refined training set = second set of data objects] which is then used to retrain the machine-learning model; the machine-learning model is tested against the validation dataset and this process continues until the process converges [i.e., until the convergence rule is met] – Sharma, paragraph 51) ….”
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wodetski/Madani/Hoque to test the model and retrain it with new training data until a criterion is met, as disclosed by Sharma, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the model is up-to-date and better geared toward fulfilling its intended purpose, thereby increasing the accuracy of the model on the task. See Sharma, paragraph 51.
Claim 9 is a method claim corresponding to system claim 2 and is rejected for the same reasons as given in the rejection of that claim. Similarly, claim 16 is a non-transitory computer-readable medium claim corresponding to system claim 2 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 3, the rejection of claim 1 is incorporated. Wodetski further discloses “determining [whether] the tested machine learning model … satisf[ies] the entity rule in accordance with the desired accuracy (Wodetski Fig. 15 shows that machine learning training is applied until a performance/accuracy crosses a desired threshold, and high-performing universal contract model rules [entity rules]/models are packaged for deployment) …; and …
[training] the tested machine learning model with the … set of data objects to meet the one or more entity rule[s] with the desired accuracy (Wodetski Fig. 15 shows that machine learning training is applied until a performance/accuracy crosses a desired threshold, and high-performing universal contract model rules [entity rules]/models are packaged for deployment).”
Wodetski/Madani/Hoque appears not to disclose explicitly the further limitations of the claim. However, Sharma discloses that “the processor is further configured to:
in response to determining that the tested machine learning model fails to satisfy the … rule in accordance with the desired accuracy:
refine one or more training rules of the set of training rules to generate a third set of data objects associated with a third set of data groups (aggregated data are then shared with data sources who use the aggregated data to train a local machine-learning model to generate new data to be provided to a central server; these new data are used to refine the training dataset [another refined training set = third set of data objects] which is then used to retrain the machine-learning model; the machine-learning model is tested against the validation dataset and this process continues until the process converges [determining that the process has not converged = determining that the model does not meet the rule that the process should converge; the fact that particular new data should be added to the training dataset may be regarded as a training rule, so that rule is refined when specific new data are added; note also that the fact that the process is iterative implies that there are multiple training datasets] – Sharma, paragraph 51; training dataset includes labels [data groups] identifying a correct classification for the datapoint – id. at paragraph 1); and
retrain the tested machine learning model with the third set of data objects to meet the one or more … rule[s] with the desired accuracy (aggregated data are then shared with data sources who use the aggregated data to train a local machine-learning model to generate new data to be provided to a central server; these new data are used to refine the training dataset [refined training set = third set of data objects] which is then used to retrain the machine-learning model; the machine-learning model is tested against the validation dataset and this process continues until the process converges [i.e., until the convergence rule/desired accuracy is met] – Sharma, paragraph 51).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wodetski/Madani/Hoque to retrain the model with new training data until a criterion is met, as disclosed by Sharma, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the model is up-to-date and better geared toward fulfilling its intended purpose, thereby increasing the accuracy of the model on the task. See Sharma, paragraph 51.
Claim 10 is a method claim corresponding to system claim 3 and is rejected for the same reasons as given in the rejection of that claim. Similarly, claim 17 is a non-transitory computer-readable medium claim corresponding to system claim 3 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 5, Wodetski, as modified by Sharma/Madani/Hoque, discloses that “the processor is further configured to refine the one or more training rules by adjusting one or more hyperparameters, adjusting sampling rules, or collecting additional historical data (aggregated data are then shared with data sources who use the aggregated data to train a local machine-learning model to generate new data to be provided to a central server; these new data [additional historical data, “historical” in the sense that they are collected prior to the training] are used to refine the training dataset which is then used to retrain the machine-learning model; the machine-learning model is tested against the validation dataset and this process continues until the process converges – Sharma, paragraph 51).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wodetski/Madani/Hoque to collect more training data for retraining the model, as disclosed by Sharma, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the model is up-to-date and better geared toward fulfilling its intended purpose, thereby increasing the accuracy of the model on the task. See Sharma, paragraph 51.
Claim 12 is a method claim corresponding to system claim 5 and is rejected for the same reasons as given in the rejection of that claim. Similarly, claim 19 is a non-transitory computer-readable medium claim corresponding to system claim 5 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 7, Wodetski, as modified by Sharma/Madani/Hoque, discloses that “the processor is further configured to collect the additional historical data by collecting augmented data or rescaling the additional historical data associated with the plurality of data groups (aggregated data are then shared with data sources who use the aggregated data to train a local machine-learning model to generate new data to be provided to a central server; these new data [additional/augmented historical data] are used to refine the training dataset which is then used to retrain the machine-learning model; the machine-learning model is tested against the validation dataset and this process continues until the process converges – Sharma, paragraph 51).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wodetski/Madani/Hoque to collect more training data for retraining the model, as disclosed by Sharma, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the model is up-to-date and better geared toward fulfilling its intended purpose, thereby increasing the accuracy of the model on the task. See Sharma, paragraph 51.
Claim 14 is a method claim corresponding to system claim 7 and is rejected for the same reasons as given in the rejection of that claim.
Claims 6, 13, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Wodetski in view of Sharma, Madani, and Hoque and further in view of Mazzoleni et al. (US 20200279171) (“Mazzoleni”).
Regarding claim 6, neither Wodetski, Madani, Hoque, nor Sharma appears to disclose explicitly the further limitations of the claim. However, Mazzoleni discloses that “the sampling rules comprise stratified sampling, cluster sampling or random sampling (various query techniques may be used to identify relevant data and non-relevant data, such as stratified sampling – Mazzoleni, paragraph 33).”
Mazzoleni and the instant application both relate to machine learning and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Wodetski, Madani, Hoque, and Sharma to use stratified sampling, as disclosed by Mazzoleni,, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that each subgroup is properly represented in the sample. See Mazzoleni, paragraph 33.
Claim 13 is a method claim corresponding to system claim 6 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 20, the rejection of claim 19 is incorporated. Sharma further discloses that “collecting the additional historical data comprises collecting augmented data or rescaling the historical data associated with the plurality of the data groups (aggregated data are then shared with data sources who use the aggregated data to train a local machine-learning model to generate new data to be provided to a central server; these new data [additional/augmented historical data] are used to refine the training dataset which is then used to retrain the machine-learning model; the machine-learning model is tested against the validation dataset and this process continues until the process converges – Sharma, paragraph 51).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wodetski/Madani/Hoque to collect more training data for retraining the model, as disclosed by Sharma, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the model is up-to-date and better geared toward fulfilling its intended purpose, thereby increasing the accuracy of the model on the task. See Sharma, paragraph 51.
Neither Wodetski, Madani, Hoque, nor Sharma appears to disclose explicitly the further limitations of the claim. However, Mazzoleni discloses that “the sampling rules comprise stratified sampling, cluster sampling or random sampling (various query techniques may be used to identify relevant data and non-relevant data, such as stratified sampling – Mazzoleni, paragraph 33) ….” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Wodetski, Madani, Hoque, and Sharma to perform stratified sampling, as disclosed by Mazzoleni, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that each subgroup is properly represented in the sample. See Mazzoleni, paragraph 33.
Response to Arguments
Applicant's arguments filed May 8, 2026 (“Remarks”) have been fully considered but they are, except insofar as rendered moot by the introduction of a new ground of rejection, not persuasive.
Applicant first argues that the claims as amended are eligible under 35 USC § 101 because (a) the testing, generating, determining, testing, and deploying limitations are allegedly not practically mentally performable and Example 39 applies; (b) the claims integrate any judicial exception recited into a practical application because the claims recite a method of avoiding threat scenarios that reduces network resources and bandwidth; and (c) the claims contain significantly more than the judicial exception and Examiner has allegedly failed to show that the claimed combination of elements is well-understood, routine, and conventional. Remarks at 11-17.
Regarding (a), but for the recitation of the use of machine learning, the generating and determining limitations are, in fact, mentally performable for the reasons given in the rejection. While Examiner agrees that training and testing the model are not practically mentally performable, and Examiner further agrees that Example 39 is applicable at least insofar as the lack of explicit recitation of mathematics renders the training and testing limitations additional elements. However, these elements nonetheless do not amount to significantly more than the judicial exception for the reasons given in the rejection.
Regarding (b), no reduction of network resources or bandwidth is claimed, and even assuming arguendo that these alleged benefits of the claimed subject matter are disclosed, which Examiner does not concede, it is unclear what nexus these alleged benefits have to the claim language itself.
Regarding (c), there is no requirement that the additional elements be shown to be well-understood, routine, and conventional in combination. Rather, the requirement that additional elements be shown to be well-understood, routine, and conventional at step 2B is invoked only when those elements amount to insignificant extra-solution activity at prong 2 of step 2A. Where necessary, Examiner has provided the appropriate Berkheimer evidence. Moreover, Applicant flatly contradicts itself in reproducing Examiner’s explicit analysis of the elements of the claim in ordered combination and then immediately proceeding to allege that Examiner has not analyzed the elements of the claim in ordered combination.
With regard to the art rejection, Applicant argues (a) modifying Wodetski with Sharma would allegedly render Wodetski inoperable for its intended purpose because Wodetski requires direct access to the full text of the documents being processed whereas the local data sources do not share their underlying data with a central server and Wodetski uses threshold-based training whereas Sharma’s convergence criterion is a statistical convergence; (b) modifying Wodetski with Sharma would allegedly change the principle of operation of Wodetski because Wodetski requires full access to the data whereas the system of Sharma does not rely on data sharing; (c) the stated motivation to combine Wodetski with Sharma is allegedly conclusory because the action allegedly fails to explain why a POSITA would look to Sharma to solve any problem recognized in Wodetski; (d) the previous action’s statement that certain aspects of the previously presented claims are ambiguous allegedly shows the weakness of the rejection; and (e) Hoque and Madani allegedly fail to teach the amended claims. Remarks at 17-24.
Arguments (a) and (b) can be jointly disposed of with the observation that Applicant’s contention that it is not making an argument that one reference must be able to be bodily incorporated into the other is disingenuous. That is precisely what Applicant is arguing. The privacy-preserving elements of Sharma, and for that matter the federated learning aspect, are in no way relied upon in the rejection. Sharma is used purely to show that testing a model in response to determining that the model satisfies a rule and deploying that model into an application were known before the effective filing date. To insist that the entire disclosure of the secondary reference must be compatible with the entire disclosure of the primary, regardless of whether those allegedly incompatible elements of the secondary reference were relied upon in the rejection itself, is to insist upon bodily incorporation as the standard for obviousness. But that is not the standard. The relevant question is not whether some aspect of Sharma that was never relied upon would change the principle of operation of Wodetski, but rather what the combined teachings of the references would suggest to an ordinary artisan. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981).
Argument (c) is another attempt by Applicant to import bodily incorporation as the standard for obviousness through the back door. Again, Examiner does not need to show that Wodetski is physically combinable with Sharma, but only needs to show that a person of ordinary skill in the art would have been motivated to modify Wodetski with the claimed features taught by Sharma before the effective filing date. Applicant does not contest the reasoning actually given in the rejection.
Argument (d) is moot because the claim language that was the subject of the note has been amended to remove the ambiguity. Nonetheless, for completeness, it should be noted that Examiner’s note in the previous Office action was a statement as to the ambiguity of the claim, and therefore a statement of how Examiner was interpreting the ambiguous element. Nothing in that note should be construed as an admission as to the strength of the rejection itself, but rather as a point of clarification with regard to claim interpretation.
Argument (e) is merely a conclusory statement. This argument fails to comply with 37 CFR 1.111(b) because it amounts to a general allegation that the claims define a patentable invention without specifically pointing out how the language of the claims patentably distinguishes them from the references. Thus, no substantive response is necessary.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 RYAN C VAUGHN whose telephone number is (571)272-4849. The examiner can normally be reached M-R 7:00a-5:00p ET.
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, Kamran Afshar, can be reached at 571-272-7796. 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.
/RYAN C VAUGHN/ Primary Examiner, Art Unit 2125
Prosecution Insights