# US 12,008,365 B2 · Systems and method for automated code analysis and tagging > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,008,365 B2 - Document ID: US12008365B2 - Status: Granted - Kind: B2 - Application number: 17/651,005 - Filed: 2022-02-14 - Published: 2023-08-17 - Granted: 2024-06-11 - Earliest claimed: 2021-06-28 - Assignees: JPMorgan Chase Bank, N.A. - Applicants: Jpmorgan Chase Bank , N.A. - Inventors: Sean Moran; Sanat Saha; Gaurav Singh; Fanny Silavong; Antonios Georgiadis; Ganesh Chandrasekar; Andy Alexander; Rob Otter; Brett Sanford - Web page: https://patentgrep.com/patent/US12008365B2 ## CPC - G06F 8/75: Structural analysis for program understanding - G06F 8/427: {Parsing} - G06F 17/16: Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Systems and methods for automated code analysis and tagging are disclosed. In one embodiment, a method for automated code analysis and tagging may include: (1) receiving, by a code annotation computer program executed by a computer processor, a training code snippet from a training codebase; (2) parsing, by the code annotation computer program, the training code snippet into a data structure; (3) quantifying, by the code annotation computer program, the data structure, (4) parsing, by the code annotation computer program, a docstring associated with the training code snippet into a plurality of keywords; (5) quantifying, by the code annotation computer program, the plurality of keywords; and (6) training, by the code annotation computer program, a code annotation model based on a similarity between the quantified data structure and a smoothing parameter for a Dirichlet prior smoothing estimate. ## Claims ### Claim 1 A method for automated code analysis and tagging, comprising: receiving, by a code annotation computer program executed by a computer processor, a training code snippet from a training codebase; parsing, by the code annotation computer program, the training code snippet into a data structure; quantifying, by the code annotation computer program, the data structure, parsing, by the code annotation computer program, a docstring associated with the training code snippet into a plurality of keywords; quantifying, by the code annotation computer program, the plurality of keywords; training, by the code annotation computer program, a code annotation model based on a similarity between the quantified data structure and a smoothing parameter for a Dirichlet prior smoothing estimate; and generating an annotation for the training code snippet by matrix multiplication of a test code-training probability matrix and a training code-word probability matrix, the test code-training code probability matrix computed from a Gaussian kernel density estimate between each test code snippet and each training code snippet from the training codebase. ### Claim 2 (depends on 1) The method of claim 1, wherein the data structure comprises an Abstract Syntax Tree (“AST”). ### Claim 3 (depends on 1) The method of claim 1, wherein the data structure comprises a Simplified Parse Tree (“SPT”). ### Claim 4 (depends on 1) The method of claim 1, wherein the data structure is quantified into a feature vector. ### Claim 5 (depends on 1) The method of claim 1, further comprising: extracting, by the code annotation computer program, structural features for a relationship between two nodes in the data structure. ### Claim 6 (depends on 1) The method of claim 1, further comprising: determining, by the code annotation computer program, a number of occurrences of the keyword in the docstring; comparing, by the code annotation computer program, the number of occurrences of the keyword in the docstring to a number of occurrences of the keyword in the training codebase; and applying, by the code annotation computer program, term frequency inverse document frequency to weight the keyword. ### Claim 7 (depends on 1) The method of claim 1, wherein the step of training the code annotation model based on a similarity between the quantified data structures and the quantified keywords comprises: tuning, by the code annotation computer program, a bandwidth of a Gaussian kernel to compute the similarity between the quantified data structures and a smoothing parameter for a Dirichlet prior smoothing estimate. ### Claim 8 (depends on 1) The method of claim 1, wherein the code annotation model comprises a probabilistic generative model. ### Claim 9 A method for automated code analysis and tagging, comprising: receiving, by a code annotation computer program executed by a computer processor, a training code snippet from a training codebase; parsing, by the code annotation computer program, the training code snippet into a data structure; quantifying, by the code annotation computer program, the data structure, parsing, by the code annotation computer program, a docstring associated with the training code snippet into a plurality of keywords; quantifying, by the code annotation computer program, the plurality of keywords; and tuning, by the code annotation computer program, a bandwidth of a Gaussian kernel to compute a similarity between the quantified data structures and a smoothing parameter for a Dirichlet prior smoothing estimate, wherein annotations for the training code snippet is generated by matrix multiplication of a test code-training code probability matrix and a training code-word probability matrix, wherein the test code-training code probability matrix is computed from a Gaussian kernel density estimate between each test code snippet and each training code snippet. ### Claim 10 A method for automated code analysis and tagging, comprising: parsing, by a code annotation computer program executed by a computer processor, an unannotated code snippet into a data structure; quantifying, by the code annotation computer program, the data structure; quantifying, by the code annotation computer program, keywords for a plurality of training code snippets by matrix multiplication of a test code-training code probability matrix and a training code-word probability matrix, wherein the test code-training code probability matrix is computed from a Gaussian kernel density estimate between each test code snippet and each of the plurality of training code snippets; computing, by the code annotation computer program, conditional probabilities of the unannotated code snippet occurring in each of the plurality of training code snippets; computing, by the code annotation computer program, a conditional probability of each of the keywords in the training code snippets for the unannotated code snippet; selecting, by the code annotation computer program, a plurality of the keywords based on the conditional probability for each keyword; and annotating, by the code annotation computer program, the unannotated code snippet with the selected keywords. ### Claim 11 (depends on 10) The method of claim 10, wherein the data structure comprises an Abstract Syntax Tree (“AST”). ### Claim 12 (depends on 10) The method of claim 10, wherein the data structure comprises a Simplified Parse Tree (“SPT”). ### Claim 13 (depends on 10) The method of claim 10, wherein the data structure is quantified into a feature vector. ### Claim 14 (depends on 10) The method of claim 10, wherein the keywords are quantified using a Dirchlet prior estimator. ### Claim 15 (depends on 10) The method of claim 10, wherein the conditional probabilities are computed using a Gaussian kernel density estimator. ### Claim 16 (depends on 10) The method of claim 10, further comprising: computing, by the code annotation computer program, a prior probability of each training code snippet. ### Claim 17 (depends on 10) The method of claim 10, further comprising: determining, by the code annotation computer program, that a prior probability of each training code snippet is constant. ### Claim 18 (depends on 10) The method of claim 10, wherein the step of annotating the unannotated code snippet with the selected keywords comprises: storing, by the code annotation computer program, the selected keywords; and associating, by the code annotation computer program, the stored selected keywords with the unannotated code snippet.