DETAILED ACTION
This final action is in response to the amendment and remarks filed on 02/03/2026 for application 17/545,358.
Claims 1, 4-5, 8, 12-13, 15, and 19-20 have been amended.
Claims 1-2, 4-9, 11-16, and 18-20 thereby remain pending in the application. Claims 1, 8, and 15 are independent claims.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
The amendment filed 02/03/2026 has been entered.
Applicant’s amendment to the claims with respect to resolving claim objections has been considered, and previous objections in the office action mailed 11/05/2025 are consequently withdrawn.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-2, 4-9, 11-16, and 18-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 1, it recites the limitation “the distances between the features of the first encoding and the features of the second encoding determined as cross entropy between labels and logits”. There is insufficient antecedent basis for the terms “labels” and “logits” in the claims; the claims do not previously describe any existing elements as “labels”, and the recited “logits” also do not appear to clearly correspond to the previously recited “logit layer” or any other previously recited claim elements. It is thereby unclear how these newly recited elements are interrelated with determining distance between “features of the first encoding” and “features of the second encoding”. Consequently, one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
For purposes of examination and as best understood in light of the specification [¶ 0032], the limitation is interpreted as “the distances between the features of the first encoding and the features of the second encoding determined as cross entropy between labels and logits, wherein the features of one of the encodings are set as labels and the features of the other encoding are set as logits”.
Regarding claims 8 and 15, they have the same deficiencies as those found in claim 1 above. Consequently, they are rejected for the same reasons as claim 1 and are likewise interpreted as detailed above.
Regarding claims 2, 4-7, 9, 11-14, 16, and 18-20, they inherit the deficiencies of their parent claims. Consequently, they are also rejected under 35 U.S.C. 112(b) as being indefinite for depending on an indefinite parent claim.
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 1-2, 4-9, 11-16, and 18-20 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”).
Independent Claims (Claim 1, Claim 8, Claim 15):
Step 1: Claim 1 is drawn to a method, claim 8 is drawn to a system/apparatus, and claim 15 is drawn to a product. Therefore, each of these claims falls under one of the four categories of statutory subject matter (process/method, machine/apparatus, manufacture/product, or composition of matter).
Step 2A Prong 1: Claims 1, 8, and 15 each recite a judicially recognized exception of an abstract idea.
Claim 1 recites, inter alia, a method comprising:
generating an explanation as to why the machine learning model classified the first instance of data and the second instance of data differently – This limitation amounts to observing two portions of text (e.g., sentences), and applying processes of reasoning (identifying difference[s]) to identify particular differences that would have contributed to the two portions of text as being classified differently. Therefore, it recites a process of evaluation capable of being performed in the human mind or using pen and paper – as an example, comparing the two sentences “global prices are on an incline” and “global temperatures are on the rise”, and determining that a difference in the words (i.e., features) prices and temperatures would have contributed to the sentences being classified differently (e.g., classified as economy versus weather), is a procedure that is capable of being performed in the human mind.
generating a first encoding associated with the first instance [representing a first set of texts]; generating a second encoding associated with the second instance [representing a second set of texts] – These limitations amount to performing generic transformations on “sets of text”, such as sentences, such that they adopt a new form – e.g., merely rephrasing or summarizing sentences. Therefore, they recite processes of evaluation capable of being performed in the human mind or using pen and paper.
learn similarities in a given pair of input objects – This limitation amounts to observing two portions of text, and applying reasoning to identify similarities between them (e.g., merely identifying two different sentences as being related to a similar concept). Therefore, it recites a process of evaluation capable of being performed in the human mind or using pen and paper.
based on the first encoding and the second encoding, identifying a difference in features of the first instance and features of the second instance, which contributed to the first instance and the second instance being classified differently by the machine learning model, a feature in the features of the first instance representing a text in the first set of texts, a feature in the features of the second instance representing a text in the second set of texts, wherein the difference represents the difference of the text in the first set of texts from the text in the second set of texts – This limitation amounts to observing two portions of text (e.g., sentences), and applying generic transformations (encoding[s]) and processes of reasoning (identifying difference[s]) to identify particular differences that would have contributed to the two portions of text as being classified differently. Therefore, they recite a process of evaluation capable of being performed in the human mind or using pen and paper – as an example, comparing the two sentences “global prices are on an incline” and “global temperatures are on the rise”, and determining that a difference in the words (i.e., features) prices and temperatures would have contributed to the sentences being classified differently (e.g., classified as economy versus weather), is a procedure that is capable of being performed in the human mind.
wherein the identifying the difference in the features of the first instance and the features of the second instance includes computing gradients, with respect to the first instance, of distances between features of the first encoding and features of the second encoding, the distances between the features of the first encoding and the features of the second encoding determined as cross entropy between labels and logits – This limitation amounts to performing a series of mathematical operations (computing gradients of calculated cross entropy values) to quantify differences between variables (features of the first instance and features of the second instance), and therefore recites mathematical calculation.
and identifying a feature having a largest negative gradient value to identify the difference in the features of the first instance and the features of the second instance, wherein the identified difference in the features are provided as the explanation – This limitation amounts to merely observing calculated gradients to identify a largest negative value as significant, and therefore recites a process of evaluation capable of being performed in the human mind or using pen and paper.
Claims 8 and 15 recite substantially similar abstract idea limitations to those recited in claim 1, and therefore the same judicial exception as claim 1.
Step 2A Prong 2: The following additional elements recited in claims 1, 8, and 15 do not integrate the recited judicial exceptions into a practical application.
Claim 1 additionally recites:
receiving a first instance of data and a second instance of data; inputting the [first/second] instance to a [first/second neural network] – These limitations amount to insignificant steps of gathering and transferring data, and are therefore insignificant extra-solution activity.
[first/second instance of data] which have been classified differently by a machine learning model trained to classify given input data – This limitation amounts to an insignificant pre-solution implementation step with regard to the gathering of input data, and is merely a tangential addition to the overall claimed procedure; it therefore recites insignificant extra-solution activity.
the first instance representing a first set of texts and the second instance representing a second set of texts; – This limitation amounts to no more than generally linking the claimed procedure to the field of use of natural language processing, as it merely limits the reach of the claimed procedure to processing text data without providing anything more.
a first neural network, the first neural network [generating]; a second neural, the second neural network [generating]; wherein the first neural network and the second neural network form a neural network architecture model trained to [learn], the neural network architecture model being a separate model from the machine learning model – These limitations amount to no more than mere instructions to apply an exception, i.e., they merely invoke neural networks as tools to perform existing abstract ideas of performing generic transformations on portions of text and identifying similarities/differences between portions of text.
each of the first neural network and the second neural network is structured to include a logit layer prior to a normalization layer, – This limitation amounts to no more than an incidental addition with regard to implementation of the neural networks, which are merely invoked as tools to perform existing abstract ideas. It therefore recites insignificant extra-solution activity.
the first encoding generated at the logit layer of the first neural network in tandem with the second encoding generated at the logit layer of the second neural network – This limitation amounts to no more than mere instructions to apply an exception. Simply stating that encodings are drawn from the raw outputs (i.e., logits) of the networks does no more merely invoke the neural networks as tools to perform an existing abstract idea of performing generic transformations on portions of text.
Claim 8 recites substantially similar additional elements to those recited in claim 1, and further recites:
A system comprising: a processor; a memory device coupled with the processor; the processor configured to at least: – This limitation amounts to mere instructions to implement an abstract idea on a computer or computer components.
Claim 15 recites substantially similar additional elements to those recited in claim 1, and further recites:
A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a device to cause the device to: – This limitation amounts to mere instructions to implement an abstract idea on a computer or computer components.
Step 2B: The additional elements recited in claims 1, 8, and 15, viewed individually or as a combination, do not provide an inventive concept or otherwise amount to significantly more than the recited abstract ideas themselves.
Claim 1 additionally recites:
receiving a first instance of data and a second instance of data; inputting the [first/second] instance to a [first/second neural network] – Receiving and transmitting data is well-understood, routine, and conventional activity (see MPEP § 2106.05(d); “Receiving or transmitting data over a network”) and therefore does not provide an inventive concept or significantly more to the recited abstract idea.
[first/second instance of data] which have been classified differently by a machine learning model trained to classify given input data – Evaluating outputted predictions of a machine learning model to provide classification explanations (e.g., interpretability/attribution methods) is well understood, routine, and conventional activity (see Linardatos et al., “Explainable AI: A Review of Machine Learning Interpretability Methods” [page 6 Interpretability Methods to Explain Deep Learning Models, page 11 Interpretability Methods to Explain any Black-Box Model]) in the field of explainable AI (XAI), and therefore does not provide an inventive concept or significantly more to the recited abstract idea
the first instance representing a first set of texts and the second instance representing a second set of texts; – Generally linking a judicial exception to the field of use of natural language processing without providing anything more does not provide an inventive concept or significantly more to the recited abstract idea.
a first neural network, the first neural network [generating]; a second neural, the second neural network [generating]; wherein the first neural network and the second neural network form a neural network architecture model trained to [learn], the neural network architecture model being a separate model from the machine learning model – Merely invoking neural networks as tools to perform existing abstract ideas does not provide an inventive concept or significantly more to the recited abstract idea.
each of the first neural network and the second neural network is structured to include a logit layer prior to a normalization layer – Using softmax activation functions, which normalize raw outputs from the final hidden layer (i.e., logits) to output a vector of probabilities, in neural networks (e.g., for classification tasks), is well-understood, routine, and conventional activity (see Siegel (“What are activation function in neural networks”) [page 9 Softmax]), and therefore does not provide an inventive concept or significantly more to the recited abstract idea.
the first encoding generated at the logit layer of the first neural network in tandem with the second encoding generated at the logit layer of the second neural network – Merely invoking neural networks as tools to perform existing abstract ideas does not provide an inventive concept or significantly more to the recited abstract idea.
Claim 8 recites substantially similar additional elements to those recited in claim 1, and further recites:
A system comprising: a processor; a memory device coupled with the processor; the processor configured to at least: – Mere instructions to implement an abstract idea on a computer or computer components do not provide an inventive concept or significantly more to the recited abstract idea.
Claim 15 recites substantially similar additional elements to those recited in claim 1, and further recites:
A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a device to cause the device to: – Mere instructions to implement an abstract idea on a computer or computer components do not provide an inventive concept or significantly more to the recited abstract idea.
Even when considered as an ordered combination, the additional elements recited in the claims ultimately do no more than place an abstract procedure of observing and analyzing model outputs to create explanations for classification differences, via mental processes and/or mathematical calculations, in the context of analyzing output of generic neural networks. As such, claims 1, 8, and 15 are not patent eligible.
Dependent Claims (Claims 2 & 4-7, Claims 9 & 11-14, Claims 16 & 18-20):
Dependent claims 2, 4-7, 9, 11-14, 16, and 18-20 narrow the scope of independent claims 1, 8, and 15, and thus merely narrow the recited judicial exceptions. With respect to the independent claims, the recited judicial exceptions are not meaningfully integrated into a practical application, and also do not amount to significantly more than the recited abstract ideas themselves. The dependent claims recite abstract idea limitations similar to those recited within the independent claims, as they also do not provide anything more than mathematical concepts or mental processes that are capable of being performed in the human mind and/or using pen and paper. The dependent claims also do not recite any further additional elements that successfully integrate the recited judicial exceptions into a practical application or amount to significantly more than the recited abstract ideas themselves. Consequently, claims 2, 4-7, 9, 11-14, 16, and 18-20 are also rejected under 35 U.S.C. 101.
Step 1: Claims 2 & 4-7 are drawn to a method, claims 9 & 11-14 are drawn to a system/apparatus, and claims 16 & 18-20 are drawn to a product. Therefore, each of these claims falls under one of the four categories of statutory subject matter (process/method, machine/apparatus, manufacture/product, or composition of matter).
Step 2A Prong 1: Claims 2, 4-7, 9, 11-14, 16, and 18-20 each recite a judicially recognized exception of an abstract idea.
Claim 2 recites the same judicial exception as claim 1.
Claim 4 recites, inter alia:
the generating further including selecting a predefined number of features having largest negative gradient values to identify the difference in the features of the first instance and the features of the second instance – This limitations amounts to merely observing calculated gradients to identify largest negative values, and therefore recites a process of evaluation capable of being performed in the human mind or using pen and paper.
Claim 5 recites, inter alia:
performing a post processing to a gradient of the computed gradients to reduce noise – This limitation amounts to merely performing a generic “processing” on to a mathematically computed value; it can thereby be reasonably interpreted as reciting a process of evaluation capable of being performed in the human mind or using pen and paper, or also as reciting further mathematical calculations.
Claim 6 recites, inter alia:
wherein the post processing includes multiplying the gradient with the first instance of data – This limitation amounts to performing further mathematical calculations (multiplying) with respect to a computed value (gradient).
Claim 7 recites, inter alia:
wherein the identifying a difference in the features of the first instance and the features of the second instance, includes providing an explanation including a ranked list of the features from the first instance and the features of the second instance, wherein the features included in the ranked list contributed to the first instance and the second instance being classified differently – This limitation amounts to providing a reasoning for features (e.g., words) identified as having contributed to the two instances (i.e., portions of text) being classified differently, including evaluating said identified features on a scale of importance. It thereby merely expands upon the same abstract idea recited in the independent claims (see Step 2A Prong 1 analysis of claim 1 above) of identifying differences between text, i.e., further recites processes of evaluation capable of being performed in the human mind or using pen and paper.
Claims 9 and 16 recite the same judicial exception as claims 8 and 15.
Claims 11-14 and 18-20 recite substantially similar abstract idea limitations to those recited in claims 4-7, and therefore recite the same judicial exceptions as claims 4-7.
Step 2A Prong 2: The following additional elements recited in claims 2, 9, and 16 do not integrate the recited judicial exceptions into a practical application.
Claim 2 additionally recites:
wherein the first neural network and the second neural network have identical hyperparameters and weights – This limitation amounts to an incidental addition with regard to implementation of the neural architecture recited in the independent claims. Wherein the independent claims merely invoke neural network architecture as a tool to perform existing abstract ideas (see Step 2A Prong 2 analysis of claim 1), this limitation does no more than further limit said architecture to being implemented as Siamese/twin networks, thereby generally linking a judicial exception to the technological environment of Siamese/twin networks without providing anything more.
Claims 9 and 16 recite substantially similar additional elements to those recited in claim 2, and therefore also do no more than generally link a judicial exception to the technological environment of Siamese/twin networks.
Step 2B: The additional elements recited in claims 2, 9, and 16, viewed individually or as a combination, do not provide an inventive concept or otherwise amount to significantly more than the recited abstract ideas themselves.
Claim 2 additionally recites:
wherein the first neural network and the second neural network have identical hyperparameters and weights – Generally linking a judicial exception to the technological environment of Siamese/twin networks without providing anything more does not provide an inventive concept or significantly more to the recited abstract idea.
Claims 9 and 16 recite substantially similar additional elements to those recited in claim 2, and therefore also do not provide an inventive concept or significantly more to the recited abstract idea.
Even when considered as an ordered combination, the additional elements recited in the claims ultimately do no more than place an abstract procedure of observing and analyzing model outputs to create explanations for classification differences, via mental processes and/or mathematical calculations, in the context of analyzing output of generic Siamese/twin networks. As such, claims 2, 4-7, 9, 11-14, 16, and 18-20 are not patent eligible.
Response to Arguments
The remarks filed 02/03/2026 have been fully considered.
Applicant’s remarks [Remarks pages 8-10] traversing the non-eligible subject matter rejections under 35 U.S.C. 101 set forth in the office action mailed 11/05/2025, in view of claims 1-2, 4-9, 11-16, and 18-20 as amended, have been considered but are not persuasive.
Applicant alleges that the limitations of the claims, when considered in combination, are integrated into a practical application (i.e., eligible under Step 2A Prong Two).
The examiner respectfully disagrees. Applicant is directed towards the grounds of rejection under 35 U.S.C. 101 with respect to amended claims 1-2, 4-9, 11-16, and 18-20 set forth above. Applicant’s arguments are further summarized and addressed below.
Applicant argues that the claims are integrated into a practical application of improving a machine learning technology to be able to provide an explanation or reasoning behind the machine learning model’s results, and cites to [¶ 0003, 0036, 00038] of the specification as providing support for the alleged improvement.
In response, the examiner notes that the judicial exception alone cannot provide the improvement (see MPEP § 2106.05(a)). Applicant’s alleged improvement of “providing an explanation behind machine learning model results” amounts to a procedure of observing and analyzing model outputs to create explanations for classification differences, which under a broadest reasonable interpretation, falls within the scope of that which is mentally performable within the human mind. As explained in the rejection above, a person would be able to compare the two sentences “global prices are on an incline” and “global temperatures are on the rise”, and predict that a difference in the words (i.e., features) prices and temperatures is what would have contributed to the sentences being classified differently (e.g., classified as economy versus weather) by a model. The fact that the given explanation is directed towards the outputted results of a machine learning model does not absolve the procedure itself of reaching said explanation from being abstract in nature.
Additionally, even assuming, for example, that observed results are obtained from a complex, “black box” model, such that explaining the reasoning behind the model’s prediction would fall outside the scope of what would be mentally performable by the human mind, it is important to note that the claim limitations that are relied upon to generate an explanation (i.e., “identifying a difference in features of the first instance and features of the second instance”) amount only to steps of mathematical calculation. The claimed procedure operates on the encodings outputted by a generic Siamese architecture via computing gradients of cross entropy measures to quantify differences, and utilizes said computed gradients as an explanation. As such, even in instances that rise beyond the scope of mental processing, the recited steps of mathematical calculation within the claims are the only claim elements that actually provide the alleged improvement, which is insufficient to support integration into a practical application.
Finally, the examiner notes that the claimed procedure does not actually result in improvement in the actual operation of a machine learning model itself, because the procedure ends with mere observation of model outputs and then outputting an identified difference. The claimed procedure does not further leverage the “identif[ied] differences” to actually modify the operation of the model itself in any specific or unconventional manner, and thereby does not rise beyond being directed towards a mere abstract concept.
Applicant has not presented further arguments with respect to the dependent claims. As such, amended claims 1-2, 4-9, 11-16, and 18-20 stand rejected under 35 U.S.C. 101.
Applicant’s remarks [Remarks pages 11-15] traversing the obviousness rejections under 35 U.S.C. 103 set forth in the office action mailed 02/03/2026, in view of claims 1-2, 4-9, 11-16, and 18-20 as amended, have been considered but are moot because the examiner has withdrawn the rejections at least in view of applicant’s amendment to the claims (e.g., newly added limitation “the distances between the features of the first encoding and features of the second encoding determined as cross entropy between labels and logits” in independent claims 1, 8 , and 15).
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 VIJAY M BALAKRISHNAN whose telephone number is (571) 272-0455. The examiner can normally be reached 10am-5pm EST Mon-Thurs.
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, JENNIFER WELCH can be reached on (571) 272-7212. 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.
/V.M.B./
Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143
Prosecution Insights