DETAILED ACTION
This action is in response to the claims filed February 26, 2026. Claims 1-20 are pending. Claims 1, 4, and 7 are independent claims. Claims 1, 3-4, and 7 have been amended.
The claim rejection under 35 U.S.C. 101 are maintained in view of Applicant’s arguments and amendments to the 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 .
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 1-20 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.
The term “high attention degree” in claims 1, 4, and 7 is a relative term which renders the claim indefinite. The term “high attention degree” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. It is not clear at what point the attention degree is considered high enough to be “high”. For the purposes of examination, the claim is interpreted to read “synthesizing a token having an attention degree among the first source codes”. Claims 2-3, 5-6, and 8-20 are rejected in view of their dependency on claims 1, 4, and 7.
The term “in which the similarity value is a value that approaches a predetermined value” in claims 1, 4, and 7 is a relative term which renders the claim indefinite. The term “approaches a predetermined value” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. It is not clear at what point the similarity is considered “approaching” a value. For the purposes of examination, the limitation is interpreted to read as having reached a predetermined threshold value. Claims 2-3, 5-6, and 8-20 are rejected in view of their dependency on claims 1, 4, and 7.
Claim 3 recites the limitation "a source code of the second learning data" in line 7. Claim 3 recites on line 4 “a source code of second learning data”. It is unclear whether the two “source codes” refer to the same source code. For the purposes of examination, claim 3 line 7 is interpreted to read as “the source code of the second learning data”.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claims 1, 4, and 7, the limitations “calculating, for a plurality of first source codes, a similarity between a first source code and a sentence explaining a specification of a desired program in a natural language” and “calculating an attention degree of each token constituting the first source code in calculation of the similarity” as drafted, are functions that, under their broadest reasonable interpretation, recite the abstract idea of a mathematical concept. The limitations “automatically generating a plurality of synthesis codes by synthesizing a token having a high attention degree among the first source codes having a similarity value, in which the similarity value is a value that approaches a predetermined value, with a second source code prepared in advance” and “generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis code” as drafted, is a function that, under its broadest reasonable interpretation, recite the abstract idea of a mental process. The limitation encompasses a human mind carrying out the function through observation, evaluation, judgement, and/or opinion, or even with the aid of pen and paper. Thus, these limitations recite and fall under the “Mathematical Concepts” and “Mental Processes” groupings of abstract ideas under Prong 1.
Under Prong 2, this judicial exception is not integrated into a practical application. The additional elements “a program generation device comprising a processor configured to execute operations”, “A method for generating a program”, and “a computer-readable non-transitory recording medium storing computer-executable program instructions that when executed by a processor cause a computer to execute operations comprising:” are recited at a high level of generality such that it amounts to no more than mere instructions to apply the exception using a generic computer, and/or mere computer components. See MPEP 2106.05(f). Accordingly, the additional elements do not integrate the recited judicial exception into a practical application and the claim is therefore directed to the judicial exception.
Under Step 2B, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements of “a program generation device comprising a processor configured to execute operations”, “A method for generating a program”, and “a computer-readable non-transitory recording medium storing computer-executable program instructions that when executed by a processor cause a computer to execute operations comprising:” amount to no more than mere instructions, or generic computer/computer components to carry out the exception. Accordingly, the claims are not patent eligible under 35 U.S.C. §101.
Regarding claims 2, 5, and 16, the limitation “wherein the calculating the similarity further comprises calculating the similarity between the sentence explaining the specification of the desired program in the natural language and the first source code and the attention degree of each token constituting the first source code …, … calculates a similarity between a sentence explaining a specification of a program in a natural language and a source code of the program and an attention degree of each token constituting the source code in calculation of the similarity” is an additional mental step. The limitation “by using a neural network, and the neural network calculates…” amounts to merely applying the generic computer component of a neural network for the abstract idea of calculating similarity and attention, which does not amount to practical application under Prong 2, not to significantly more under Step 2B. See MPEP 2106.05(f).
Claims 3, 6, and 17 do not recite additional mental steps. The limitation “training the neural network such that the similarity calculated by the neural network for a sentence of first learning data and a source code of second learning data in a set of learning data including a set of a source code of a program and a sentence explaining a specification of the program in a natural language approaches a predetermined similarity between a source code of the first learning data and a source code of the second learning data” amounts to mere instructions to apply training using a generic computer, which does not amount to practical application under Prong 2, nor to significantly more under Step 2B. See MPEP 2106.05(f).
Regarding claims 8, 12, and 18, no additional mental steps are recited. The limitation “wherein the neural network represents a similarity calculation model, and the similarity calculation model includes a first set of neural networks to generate a first vector based on each word in the sentence” merely applies a generic computer component of a neural network to the judicial exception of generating a vector based on words in a sentence, which does not amount to practical application under Prong 2, nor to significantly more under Step 2B. See MPEP 2106.05(f).
Regarding claims 9, 14, and 19, no additional mental steps are recited. The limitation “wherein the similarity calculation model further includes a second set of neural networks to generate a second vector based on each token in the first source codes” merely applies a generic computer component of a neural network to the judicial exception of generating a vector based on tokens in a source code, which does not amount to practical application under Prong 2, nor to significantly more under Step 2B. See MPEP 2106.05(f).
Regarding claims 10, 15, and 20, the limitation “wherein the similarity is based on a cosine similarity between the first vector and the second vector” recites an additional mental step. The same generic computer/computer components are recited as in claims 1, 4, and 7 which do not amount to practical application under Prong 2, nor to significantly more under Step 2B as discussed above.
Regarding claims 11 and 13, the limitation “wherein the automatically generated plurality of synthesis codes represents an executable code” merely describes the “automatically generated plurality of synthesis codes” of the mental step of claim 1. Therefore, claims 11 and 13 amount to additional mental steps.
Claim Rejections - 35 USC § 103
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 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 4-5, 7-10, 12, 14-16, and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over “Multi-modal Attention Network Learning for Semantic Code Retrieval” by Wan et. al (hereinafter “Wan”), in view of “Program Synthesis: Challenges and Opportunities” by David and Kroening (hereinafter “David”), further in view of “Deep Code Comment Generation” by Hu et. al (hereinafter “Li”, as consistent with usage in the non-final office action filed November 26, 2025.), and further in view of US 20220012020 A1 (hereinafter “Barke”).
Regarding claim 1, Wan disclose:
A program generation device comprising a processor configured to execute operations comprising:
- calculating, for a plurality of first source codes, a similarity between a first source code and a sentence explaining a specification of a desired program in a natural language (Page 19, “In other words, given an arbitrary code snippet x and an arbitrary description d, we want it to predict a high similarity if d is a correct description of x, and a small similarity otherwise. In training phase, we construct each training instance as a triple < x, d+, d− >: for each code snippet x, there is a positive description d+ (a correct description of x) as well as a negative description (an incorrect description of x) d− randomly chosen from the pool of all d+’s. When trained on the set of < x, d+, d− > triples, the MMAN predicts the cosine similarities of both < x, d+ > and < x, d− > pairs and minimizes the hinge range loss”) [Examiner’s remarks: For each of a plurality of source codes (for each code snippet x), a cosine similarity is predicted between x (the code) and d+ (a correct description of x).];
- calculating an attention degree of each token constituting the first source code in calculation of the similarity (Page 18, “For tokens, not all tokens contribute equally to the final semantic representation of code snippet. Therefore, we introduce the attention mechanism on tokens to extract the ones that are more important to the representation of a sequence of code tokens. The attention score for tokens
α
t
o
k
is calculated as follows”) [Examiner’s remarks: For each token in the token sequence, an attention score is calculated to indicate its importance for future operations.];
Wan does not explicitly disclose:
- automatically generating a plurality of synthesis codes by synthesizing a token having a high attention degree among the first source codes having a similarity value, in which the similarity value is a value that approaches a predetermined value, with a second source code prepared in advance; and
- generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes.
However, David discloses:
- automatically generating a plurality of synthesis codes (Page 2, “Automated program synthesis has the potential to change the way general-purpose computational devices are used by enabling non-expert users to solve problems in an automated fashion without designing and implementing a new algorithm. Essentially, program synthesis generates an implementation of the program that satisfies a given correctness specification. Consequently, it enables the workflow illustrated in figure 2,where, in order to program a computer to solve a given problem, a user only needs to give a specification of the expected result to the program synthesizer”; Page 6, “To bootstrap GP in the first iteration of the CEGIS loop, a population of random programs needs to be generated. Then, this population is iteratively evolved by applying the genetic operators CROSSOVER and MUTATE”) [Examiner’s remarks: A plurality of synthesis codes are generated.]…
- generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes (Page 2, “Automated program synthesis has the potential to change the way general-purpose computational devices are used by enabling non-expert users to solve problems in an automated fashion without designing and implementing a new algorithm. Essentially, program synthesis generates an implementation of the program that satisfies a given correctness specification. Consequently, it enables the workflow illustrated in figure 2,where, in order to program a computer to solve a given problem, a user only needs to give a specification of the expected result to the program synthesizer [generating a plurality of outputs that satisfies the specification of the desired program]”; Page 6, “The final strategy is genetic programming (GP) [26,27]. GP provides an adaptive way of searching through the space of programs for an individual that is ‘fit’ in some sense. Commonly, the fitness of an individual is measured by counting the number of tests in INPUTS for which it satisfies the specification. To bootstrap GP in the first iteration of the CEGIS loop, a population of random programs needs to be generated. Then, this population is iteratively evolved by applying the genetic operators CROSSOVER and MUTATE. CROSSOVER combines selected existing programs into new programs, whereas MUTATE randomly changes parts of a single program. Fitter programs are more likely to be selected [wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes]”) [Examiner’s remarks: A plurality of synthesis codes are generated and tested for fitness (whether they fit a specification as defined by tests). If they fail, they are less likely to be randomly selected.].
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teachings of David into the teachings of Wan to include “automatically generating a plurality of synthesis codes” and “generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes”. As stated in David, “Automated program synthesis has the potential to change the way general-purpose computational devices are used by enabling non-expert users to solve problems in an automated fashion without designing and implementing a new algorithm” (Page 2). Automating program synthesis makes programming more accessible to more people. David describes known methods of performing program synthesis, including genetic algorithms which perform random search. Therefore, it would be obvious to one of ordinary skill in the art to combine code generation with automated synthesis of code through random selection.
The combination of Wan and David does not explicitly disclose:
- … by synthesizing a token having a high attention degree among the first source codes having a similarity value, in which the similarity value is a value that approaches a predetermined value, with a second source code prepared in advance; and
However, Li discloses:
- … by synthesizing a token having a high attention degree among the first source codes (Page 203, “Attention mechanism is a recent model that selects the important parts from the input sequence for each target word. For example, the token “whether” in comments usually aligns with the “if” statements in the source code. The generation of each word is guided by a classic attention method proposed by Bahdanau et al”) [Examiner’s remarks: Li discloses choosing tokens for a generated sequence based on high attention indicating alignment. Wan discloses calculating attention. Therefore, one of ordinary art in the skill may apply the attention calculations of Wan to the usage indicated by Li.])
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teachings of Li into the combined teachings of Wan and Hu to include “by synthesizing a token having a high attention degree among the first source codes”. As stated in Li, “Attention mechanism is a recent model that selects the important parts from the input sequence for each target word” (Page 203). Attention allows a model to indicate important parts of input sequences and allows for longer inputs without the same loss of context. Selecting words based on importance to an input sequence ensures that the output sequence is related to the task provided by the input. Therefore, it would be obvious to one of ordinary skill in the art to combine code generation with token selection based on attention score.
The combination of Wan, David, and Li does not explicitly disclose:
- having a similarity value, in which the similarity value is a value that approaches a predetermined value, with a second source code prepared in advance; and
However, Barke discloses:
- having a similarity value, in which the similarity value is a value that approaches a predetermined value, with a second source code prepared in advance (Paragraph [0033], “A feedback engine analyzes the additional inputs 118, 120 to classify them as either a positive input or a negative input. The positive inputs represent locations where the transformation should produce an edit and a negative input represents where the transformation should not produce an edit”; Paragraph [0061], “The feedback engine 322 categorizes an additional input based on the reword score and the thresholds (block 520). An additional input is classified as a positive input when the reword score is greater than the positive input threshold, p, and an additional input is classified as a negative input when the reward score is less than the negative input threshold, n (block 520)”; Paragraph [0081], “The reward score for each input that includes the cursor position is set using a distance function that is based on a code-similarity metric between an input AST and an example input AST (e.g., clone-detection based code similarity metrics or tree edit distance)”) [Examiner’s remarks: Barke discloses using a similarity to an example input AST (source code prepared in advance) to determine if the input is positive based on a similarity threshold (positive input threshold). The threshold is used to determine a program transformation.]; and
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teachings of Barke into the combined teachings of Wan, Hu, and Li to include “having a similarity value, in which the similarity value is a value that approaches a predetermined value, with a second source code prepared in advance”. As stated in Barke, “Program synthesis is a technique that learns a program in a programming language that meets a developer's intent as expressed in some specification. A goal is to generate a program that is consistent with the provided examples and to produce outputs on all additional positive inputs and not on any additional negative inputs” (Paragraph [0021]). Automated program synthesis allows for a lower level of entry into coding by allowing users with less experience to obtain programs based only on specifications. Identifying similar programs allows for identification of previous iterations of a similar problem, or identification of when a generated code is similar to an intended solution. Therefore, it would be obvious to one of ordinary skill in the art to combine code generation with similarity analysis.
Regarding claim 2, the rejection of claim 1 is incorporated; and Wan further discloses:
- wherein the calculating the similarity further comprises calculating the similarity between the sentence explaining the specification of the desired program in the natural language and the first source code and the attention degree of each token constituting the first source code by using a neural network, and the neural network calculates a similarity between a sentence explaining a specification of a program in a natural language and a source code of the program and an attention degree of each token constituting the source code in calculation of the similarity (Page 18, “For tokens, not all tokens contribute equally to the final semantic representation of code snippet. Therefore, we introduce the attention mechanism on tokens to extract the ones that are more important to the representation of a sequence of code tokens [and the attention degree of each token constituting the first source code… and an attention degree of each token constituting the source code]”; Page 19, “When trained on the set of < x, d+, d− > triples, the MMAN predicts the cosine similarities of both < x, d+ > and < x, d− > pairs and minimizes the hinge range loss [wherein the calculating the similarity further comprises calculating the similarity between the sentence explaining the specification of the desired program in the natural language and the first source …by using a neural network, and the neural network calculates a similarity between a sentence explaining a specification of a program in a natural language and a source code of the program … in calculation of the similarity]”) [Examiner’s remarks: The neural network calculates similarity between a specification of a program (d+) and the desired program (x). It calculates attention for each token to determine which should be more important in the later calculations.].
Regarding claim 8, the rejection of claim 1 is incorporated; and Wan further discloses:
- wherein the neural network represents a similarity calculation model, and the similarity calculation model includes a first set of neural networks to generate a first vector based on each word in the sentence (Page 19, “In the training phase, the descriptions are extracted from the code comments, while in the testing phase, the description are regarded as the input queries. In this paper, we apply a vallina LSTM to represent the description. … where i = 1,..., |d| and w is the word embedding layer to embed each word into a vector. The hidden state of last step
h
|
d
|
d
e
s
can be used as a vector representation of d.”) [Examiner’s remarks: A vector is generated to represent the description d of a desired code.].
Regarding claim 9, the rejection of claim 8 is incorporated; and Wan further discloses:
- wherein the similarity calculation model further includes a second set of neural networks to generate a second vector based on each token in the first source codes (Page 17, “We apply a LSTM to represent the tokens of code, and a Tree-LSTM to represent the AST of code a GGNN to represent the CFG of code”, “In this paper, we apply LSTM network to represent the sequential tokens…where i = 1,..., |x|, w is the word embedding layer to embed each word into a vector. The final hidden state htok |x| of the last token of code is the token modality representation of x”) [Examiner’s remarks: A vector is generated based on the tokens of a source code using a different LSTM.].
Regarding claim 10, the rejection of claim 9 is incorporated; and Wan further discloses:
- wherein the similarity is based on a cosine similarity between the first vector and the second vector (Page 16, “While considering the code snippet x and description d, since these two modalities are from different sources, it is desirable for us to apply coordinated representation for them… Examples of such coordination include minimizing cosine distance, or maximizing correlation. In this paper, the cosine similarity function is adopted”; Page 19, “When trained on the set of < x, d+, d− > triples, the MMAN predicts the cosine similarities of both < x, d+ > and < x, d− > pairs and minimizes the hinge range loss”) [Examiner’s remarks: Wan discloses calculating a cosine distance (cosine similarity) for the distance between a description and a code.].
Claims 4-5, 12, and 14-15 are method claims corresponding to the system claims hereinabove (claims 1-2 and 8-10, respectively). Therefore, claims 4-5, 12, and 14-15 are rejected for the same reasons set forth in the rejections of claims 1-2 and 8-10, respectively.
Claims 7, 16, and 18-20 are computer-readable storage medium claims corresponding to the system claims hereinabove (claims 1-2 and 8-10, respectively). Therefore, claims 7, 16, and 18-20 are rejected for the same reasons as set forth in the rejections of claims 1-2 and 8-10 respectively.
Claims 3, 6, and 17 are rejected under 35 U.S.C. 103 as being unpatentable over “Multi-modal Attention Network Learning for Semantic Code Retrieval” by Wan et. al (hereinafter “Wan”), in view of “Program Synthesis: Challenges and Opportunities” by David and Kroening (hereinafter “David”), further in view of “Deep Code Comment Generation” by Hu et. al (hereinafter “Li”, as consistent with usage in the non-final office action filed November 26, 2025.), and further in view of US 20220012020 A1 (hereinafter “Barke”), further in view of “DeepSim: Deep Learning Code Functional Similarity” by Zhao et. al (hereinafter “Zhao”).
Regarding claim 3, the rejection of claim 2 is incorporated; and Wan further discloses:
- training the neural network such that the similarity calculated by the neural network for a sentence of first learning data and a source code of second learning data in a set of learning data including a set of a source code of a program and a sentence explaining a specification of the program in a natural language approaches a predetermined similarity (Page 16, “While considering the code snippet x and description d, since these two modalities are from different sources, it is desirable for us to apply coordinated representation for them [6], which is defined as follows: … where each modality has a corresponding projection function (g1 and g2 above) that projects it into an intermediate semantic space with a similarity constraint/coordination on them. Examples of such coordination include minimizing cosine distance, or maximizing correlation. In this paper, the cosine similarity function is adopted.”; Page 19, “When trained on the set of < x, d+, d− > triples, the MMAN predicts the cosine similarities of both < x, d+ > and < x, d− > pairs and minimizes the hinge range loss [6], [51]: L(θ) =
∑
<
x
,
d
+
,
d
-
>
∈
D
m
a
x
(
0
,
β
-
s
i
m
x
,
d
+
+
s
i
m
x
,
d
-
)
, where θ denotes the model parameters, D denotes the training dataset, sim denotes the similarity score between code and description β is a small constant margin. x, d+ and d− are the embedded vectors of x, d+ and d−, respectively. In our experiments, we adopt the cosin similarity function (cf. IV-F) and set the fixed β value to 0.05. Intuitively, the ranking loss encourages the similarity between a code snippet and its correct description to go up, and the similarities between a code snippet and incorrect descriptions to go down”) [Examiner’s remarks: Wan discloses finding a similarity score between a description of the code and a code, and training until the difference is minimized.]…
The combination of Wan, David, Li, and Barke does not explicitly disclose:
- …between a source code of the first learning data and a source code of the second learning data.
However, Zhao discloses:
- …between a source code of the first learning data and a source code of the second learning data (Abstract, “Measuring code similarity is fundamental for many software engineering tasks, e.g., code search, refactoring and reuse. However, most existing techniques focus on code syntactical similarity only, while measuring code functional similarity remains a challenging problem. In this paper, we propose a novel approach that encodes code control flow and data flow into a semantic matrix in which each element is a high dimensional sparse binary feature vector, and we design a new deep learning model that measures code functional similarity based on this representation”) [Examiner’s remarks: Zhao discloses finding similarity between two pieces of code for code search, refactoring and reuse. Comparing distance to the distance between two codes of similar function gives the model an idea of when two things of similar function but not syntax are compared.].
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teachings of Zhao into the combined teachings of Wan, David, Li, and Barke to include “between the source code of the first learning data and the source code of the second learning data”. As stated in Zhao, “Measuring code similarity is fundamental for many software engineering tasks, e.g., code search, refactoring and reuse” (Abstract). Comparing code and description similarity to code similarity of two codes gives the model an idea of when code with similar function but not syntax are close enough. Therefore, it would be obvious to one of ordinary skill in the art to combine code generation with comparing to the similarity of two source codes.
Claim 6 is a method claim corresponding to the system claim hereinabove (claim 3). Therefore, claim 6 is rejected for the same reasons set forth in the rejection of claim 3.
Claim 17 is a computer-readable storage medium claim corresponding to the system claim hereinabove (claim 3). Therefore, claim 17 is rejected for the same reasons set forth in the rejection of claim 3.
Claims 11 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over “Multi-modal Attention Network Learning for Semantic Code Retrieval” by Wan et. al (hereinafter “Wan”), in view of “Program Synthesis: Challenges and Opportunities” by David and Kroening (hereinafter “David”), further in view of “Deep Code Comment Generation” by Hu et. al (hereinafter “Li”, as consistent with usage in the non-final office action filed November 26, 2025.), and further in view of US 20220012020 A1 (hereinafter “Barke”), and further in view of “Code Generation from Supervised Code Embeddings” by Hu et. al (hereinafter “Hu”).
Regarding claim 11, the rejection of claim 1 is incorporated; and the combination of Wan, David, Li, and Barke does not explicitly disclose:
- wherein the automatically generated plurality of synthesis codes represents an executable code.
However, Hu discloses:
- wherein the automatically generated plurality of synthesis codes represents an executable code (Page 389, “Some researchers utilize a standard or a variant encoder-decoder model to map NL to a snippet of executable code directly”).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the teachings of Hu into the combined teachings of Wan, David, Li, and Barke to include “wherein the automatically generated plurality of synthesis codes represents an executable code”. As stated in Hu, “Traditional code generation approaches are usually based on matching similar code snippets” (Page 389). Automated generation of executable code allows for faster code testing and deployment during the development process. Code matching via encoder-decoder to generate executable code is a known method for code generation. Therefore, it would be obvious to one of ordinary skill in the art to combine code generation with automatic generation of executables.
Claim 13 is a method claim corresponding to the system claims hereinabove (claim 11). Therefore, claim 13 is rejected for the same reasons as set forth in the rejection of claim 11.
Response to Arguments
Applicant’s arguments with respect to claims 1-20 under 35 U.S.C. 103 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
In the remarks, Applicant argues:
The Office Action asserted that claim 1 falls within the 'Mental Process' grouping of abstract ideas." (Office Action, at page 4).
Applicant respectfully disagrees with the assertion. Amended claim 1, under its broadest interpretation, is directed to a technical solution to solve the technical problem of providing methods for improve a probability of generating a desired program. According to MPEP § 2106.04(a)(2)(III)(A), "A Claim With Limitation(s) That Cannot Practically be Performed in the Human Mind Does Not Recite a Mental Process. Claims do not recite a mental process when they do not contain limitations that can practically be performed in the human mind, for instance when the human mind is not equipped to perform the claim limitations. See ... SiRF Tech., Inc. v. Int'l Trade Comm'n, 601 F.3d 1319, 94 USPQ2d 1607 (Fed. Cir. 2010), as directed to inventions that 'could not, as a practical matter, be performed entirely in a human's mind')." (Emphasis added).
Amended claim 1 recites the limitations of "generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes." (Emphasis added). At least the above limitations cannot be practically performed in the human mind. For example, a human mind cannot practically perform "repeatedly replacing program tokens with pre-generated program based on the outcome of the code." Applying the rule in MPEP § 2106.04(a)(2)(III)(A), claim 1 does not fall into the grouping of mental process.
Therefore, Applicant respectfully submits that amended claim 1 does not fall into the grouping of a mental process as asserted in the Office Action.
[See Remarks – Pages 8-9]
Examiner’s Response:
Examiner respectfully disagrees. The limitation of “generating a plurality of outputs that satisfies the specifications of the desired program” can be performed by the human mind, because the human mind is capable of creating programs which satisfies a specification. In regards to applicant’s specific argument that “repeatedly replacing program tokens with pre-generated program based on the outcome of the code” can not be done by the human mind, a human mind is able to select pre-generated programs and replace tokens in a program with them, with the aid of pen and paper. The human mind may also do this repeatedly in response to a given outcome of the code. Therefore, the limitation is directed to a mental process.
In the remarks, Applicant argues:
Even assuming, arguendo, amended claim 1 falls into the grouping of a mental process, amended claim 1 is still not directed to an abstract idea because amended claim 1 as a whole integrates the alleged judicial exceptions into a practical application (e.g., "generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes").
According to MPEP § 2106.04(d)(1), "[a] claim reciting a judicial exception is not directed to the judicial exception if it also recites additional elements demonstrating that the claim as a whole integrates the exception into a practical application. One way to demonstrate such integration is when the claimed invention improves the functioning of a computer or improves another technology or technical field." (Emphasis added).
Applying the rule set forth in MPEP § 2106.04(d)(1), amended claim 1 recites specific improvements to the technical field of providing methods for improve a probability of generating a desired program. For example, amended claim 1 recites limitations of "generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes." (Emphasis added). Auto-generation of codes will provide efficiency in both productivity and cost associated with the IT.
Additionally, these data analysis steps are different from "a claim to 'collecting information, analyzing it, and displaying certain results of the collection and analysis,' where the data analysis steps are recited at a high level of generality such that they could practically be performed in the human mind, Electric Power Group, LLC v. Alstom, S.A." (ad.; 48 8309 F.3d 1350, 1356 (Fed. Cir. 2016); emphasis added). Contrarily to Electric Power Group, amended claim 1 recites detailed features (e.g., "generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes" etc.), which are beyond a high-level of generality.
MPEP § 2106.05(a) further explains that "the claim must be evaluated to ensure the claim itself reflects the disclosed improvement in technology . . . In making this determination, it is critical that examiners look at the claim 'as a whole,' in other words, the claim should be evaluated 'as an ordered combination, without ignoring the requirements of the individual steps.' When performing this evaluation, examiners should be 'careful to avoid oversimplifying the claims' by looking at them generally and failing to account for the specific requirements of the claims. McRO, 837 F.3d at 1313, 120 USPQ2d at 1100." (Emphasis added). Applying the rules of MPEP here, claim 1 recites limitations of "generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes."
Therefore, Applicant respectfully submits that amended claim 1 is not directed to an abstract idea even if one were to assume that claim 1 falls into the grouping of a mental process.
[See Remarks – Pages 9-10]
Examiner’s remarks:
Examiner respectfully disagrees. The cited limitation “generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes” is not rejected for its failure to integrate the abstract idea into a practical application under Prong 2 of the Alice framework. Rather, the claim is rejected under Prong 1 for being directed to the abstract idea of a mental process. Therefore, the limitation is rejected under Prong 1, and the analysis does not move on to Prong 2 for this limitation.
In the remarks, Applicant argues:
Even assuming, arguendo, amended claim 1 is directed to the judicial exception of an abstract idea, amended claim 1 is patent eligible because it recites additional elements that are "unconventional or otherwise more than what is well-understood, routine, conventional activity in the field." (Section III(B) of 2019 PEG.).
According to MPEP § 2106.05(a), "[a]n important consideration in determining whether a claim improves technology is the extent to which the claim covers a particular solution to a problem or a particular way to achieve a desired outcome" and "the improvement can be provided by the additional element(s) in combination with the recited judicial exception." (Emphasis added).
Applying the rule set forth in MPEP § 2106.05(a), amended claim 1 recites a particular solution to address the computer-centric challenge of providing methods for improve a probability of generating a desired program. For example, claim 1 recites "generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes." (Emphasis added).
These features are neither well-understood, routine, nor conventional in the field.
Further, according to MPEP § 2106.05, "[l]imitations that the courts have found to qualify as "significantly more" when recited in a claim with a judicial exception include: .. . Applying the judicial exception with, or by use of, a particular machine, e.g., a Fourdrinier machine (which is understood in the art to have a specific structure comprising a headbox, a paper-making wire, and a series of rolls) that is arranged in a particular way to optimize the speed of the machine while maintaining quality of the formed paper web, as discussed in Eibel Process Co. v. Mnn. & Ont. Paper Co., 261 U.S. 45, 64-65 (1923) (see MPEP § 2106.05(b))." (MPEP § 2106.05). Similarly, here, claim 1 recites and uses a particular machine, such as machine learning component. Accordingly, the limitations recited in claim 1 qualifies as "significantly more".
In addition to the foregoing, Applicant respectfully notes the recent developments on the subject matter eligibility stated on the Memorandum dated August 4, 2025, which highlights:
"The mental process grouping is not without limits. Examiners are reminded not to expand this grouping in a manner that encompasses claim limitations that cannot practically be performed in the human mind. The MPEP and the AI-SME Update provide examples of claim limitations that cannot be practically performed in the human mind. Claim limitations that encompass AI in a way that cannot be practically performed in the human mind do not fall within this grouping. Distinguishing claims that recite a judicial exception from claims that merely involve a judicial exception: Examiners should be careful to distinguish claims that recite an exception (which require further eligibility analysis) from claims that merely involve an exception (which are eligible and do not require further eligibility analysis)." Memorandum, pages 2-3 (emphasis in the original).
In summary, amended claim 1 is directed to patent-eligible subject matter, because amended claim 1 recites features that do not fall into one of the enumerated groupings set forth in the 2019 PEG; and/or, because the features are integrated into a practical application; and/or, alternatively because amended claim 1 recites additional elements that are not well- understood, routine, or conventional in the field. Accordingly, Applicant respectfully requests reconsideration and withdrawal of the 35 U.S.C. § 101 rejections of amended claim 1 and its dependent claims. Similar arguments also apply to amended claims 4 and 7 and respective dependent claims.
Therefore, Applicant respectfully submits that the rejections under 35 U.S.C. § 101 have been overcome and should be withdrawn.
Withdrawal of the § 101 rejections is therefore respectfully requested.
Examiner’s response:
Examiner respectfully disagrees. The cited limitation “generating a plurality of outputs that satisfies the specification of the desired program, wherein if the generated output does not satisfy the specification, a plurality of synthesis codes is generated by randomly selecting one or more pre-generated program components and replacing the token of the synthesis codes” is not rejected for its failure to amount to significantly more under Step 2B of the Alice framework. Rather, the claim is rejected under Prong 1 for being directed to the abstract idea of a mental process. Therefore, the limitation is rejected under Prong 1, and the analysis does not move on to Prong 2 for this limitation.
In regards to Applicants argument that “claim 1 recites and uses a particular machine, such as machine learning component”, neither claims 1, 4, nor 7 recite use of a specific machine learning component, nor does it recite the general use of a machine learning program. Rather the claims recite use of a “program generation device comprising a processor” and “computer-readable, non-transitory computing medium” which are generic computer components which does not amount to significantly more than the abstract idea. See MPEP 2106.05(f). In regards to Applicant’s argument directed towards the Memorandum dated August 4, 2025, the claims as currently recited do not use AI in a way that cannot be practically performed in the human mind. The claims recite calculating similarity between two entities, calculating an attention degree, generating code satisfying certain parameters, and generating code by randomly selecting pre-generated program components to replace existing tokens under certain conditions, all of which may be practically performed in the human mind, with the aid of pen and paper. Therefore, the rejections of the claims under 35 U.S.C. 101 is proper, and maintained.
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 VIVIAN WEIJIA DUAN whose telephone number is (703)756-5442. The examiner can normally be reached Monday-Friday 8:30AM-5PM.
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, Wei Y Mui can be reached at (571) 272-3708. 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.W.D./Examiner, Art Unit 2191
/WEI Y MUI/Supervisory Patent Examiner, Art Unit 2191
Prosecution Insights