Prosecution Insights
Last updated: July 17, 2026
Application No. 18/219,763

ENCODING LOG-SPECIFIC NUMERICAL LEXICAL TOKENS WITH SUBSTITUTE NUMERIC LEXICAL TOKENS

Non-Final OA §103
Filed
Jul 10, 2023
Examiner
LERNER, MARTIN
Art Unit
2658
Tech Center
2600 — Communications
Assignee
ORACLE INTERNATIONAL Corporation
OA Round
4 (Non-Final)
78%
Grant Probability
Favorable
4-5
OA Rounds
0m
Est. Remaining
91%
With Interview

Examiner Intelligence

Grants 78% — above average
78%
Career Allowance Rate
772 granted / 990 resolved
+16.0% vs TC avg
Moderate +13% lift
Without
With
+13.4%
Interview Lift
resolved cases with interview
Typical timeline
2y 11m
Avg Prosecution
25 currently pending
Career history
1016
Total Applications
across all art units

Statute-Specific Performance

§101
9.9%
-30.1% vs TC avg
§103
74.2%
+34.2% vs TC avg
§102
3.5%
-36.5% vs TC avg
§112
8.8%
-31.2% vs TC avg
Black line = Tech Center average estimate • Based on career data from 990 resolved cases

Office Action

§103
CTFR 18/219,763 CTFR 76255 DETAILED ACTION 07-03-aia AIA 15-10-aia 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 § 103 07-20-aia AIA 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. 07-21-aia AIA Claim s 1, 3 to 7, 13 to 14, 16, 18, 20, and 23 are rejected under 35 U.S.C. 103 as being unpatentable over Bhatia et al. (U.S. Patent Publication 2021/0026722) in view of Moore et al. (U.S. Patent Publication 2021/0174253) . Concerning independent claims 1 and 16, Bhatia et al. discloses a method and computer program product for detecting an anomaly in an event log, comprising: “extracting a first original numeric literal and a second original numeric literal from a log entry from a plurality of log entries in a log” – a method identifies anomalies in received monitoring logs from an endpoint log source (Abstract); data is log data that describes attributes of particular systems including connection log data; exemplary connection log data is made up of over one million logs each of which have multiple, e.g. , 20 or more, features, i.e. , attributes of the system that is being monitored (¶[0054]); an original data set 202 shows five data logs, i.e. , records of features, of multiple systems; that is Log 0 is for 15 parameters from a first set of systems, Log 1 is for these same 15 parameters but for a second set of systems, etc . (¶[0064]: Figure 2); Figure 2 illustrates “a plurality of log entries in a log” for Log 0 to Log 4; a parameter of ‘46131’ for total bytes is “a first original numeric literal” and a parameter of ‘443’ for port is “a second original numeric literal” of “a log entry” represented by Log 0; “generating a first substitute numeric literal [and a third substitute literal that together] represent the first original numeric literal and a second substitute numeric literal that represents the second original numeric literal” – dataset 301 shows the data from original data set 202 after it has been transformed in order to show both the 15 parameters from the original data set 202 as well as 10 new parameters that are created by evaluating the original data set 202 (¶[0065]: Figure 3); Figure 3 illustrates “a first substitute numeric literal” of 0.22423 that represents “the first original numeric literal” of total bytes and “a second substitute numeric literal” of -0.96799 that represents “the second original numeric literal” of port in “a log entry” of Log 0; “wherein the first substitute numeric literal does not represent the second original numeric literal, and the log entry does not contain the first substitute numeric literal and the second substitute numeric literal” – here, “the first substitute numeric literal” of 0.22423 only represents “the first original numeric literal” and does not represent “the second original numeric literal” (“wherein the first substitute numeric literal does not represent the second original numeric literal”); similarity, Log 0 is “the log entry” that does not contain “the first substitute numeric literal” of 0.22423” or “the second substitute numeric literal” of -0.96799 (“the log entry does not contain the first substitute numeric literal and the second substitute numeric literal”); “generating a sequence of lexical tokens that represents the log entry and contains the first substitute numeric literal[, the third substitute numeric literal,] and the second substitute numeric literal, wherein the sequence of lexical tokens does not contain the first original numeric literal and the second original numeric literal” – dataset 301 shows the data from original data set 202 after it has been transformed in order to show both the 15 parameters from the original data set 202 as well as 10 new parameters that are created by evaluating the original data set 202 (¶[0065]: Figure 3); Figure 3 illustrates “a sequence of lexical tokens that represents the log entry” in a leftmost numerical column of the table which represents Log 0, and contains “the first substitute numeric literal” of 0.22423 for total bytes and “the second substitute numeric literal” of -0.96799 for port; “a sequence of lexical tokens” represented by the leftmost numerical column of the table in Figure 3 does not contain “the first original numeric literal” of 46131 for total bytes or “the second original numeric literal” of 443 for port of Figure 2 (“wherein the sequence of lexical tokens does not contain the first original numeric literal and the second original numeric literal”); “generating, by a machine learning model and based on the sequence of lexical tokens that represents the log entry and contains the first substitute numeric literal[, the third substitute numeric literal,] and the second substitute numeric literal, an inference that characterizes the log entry” – anomaly scoring logic scores how ‘certain’ the system is that a particular anomaly is in fact anomalous, i.e. , shows a feature of the monitored system that is not found in normal/error-free operations (¶[0062]); data from dataset 404 is fed into an unsupervised machine learning logic; any anomaly score that is over 0.0, i.e. , a positive score, is deemed to be normal, but any anomaly score that is negative is deemed to be potentially anomalous (¶[0067]: Figure 4); once the security information and management system (STEM) 626 has evaluated the high-priority anomalous logs, it sends the evaluated anomalous logs to a log confirmation logic (LCL) 628, which utilizes machine learning 625 to confirm that the high-priority anomalous logs actually contain anomalous log data (¶[0081]: Figure 6). Concerning independent claims 1 and 16, Bhatia et al. discloses all of the limitations with an arguable exception of generating “a third substitute numeric literal that together” with a first substitute numeric literal represent the first original numeric literal. That is, Bhatia et al. clearly discloses that a first substitute numeric literal of 0.22423 represents a first original numeric literal for total bytes, but not that there is an additional “third substitute numeric literal” that represents the first original numeric literal. Still, Bhatia et al. actually discloses that dataset 301 of substitute numeric literals of Figure 3 includes both the 15 parameters from the original data set as well as 10 new parameters that are created by evaluating the original data set 202. (¶[0065]: Figure 3) Consequently, Figure 3 shows additional substitute numeric literals so that there are 25 of them as compared to 15 of the original numeric literals. Arguably, Bhatia et al. discloses that “a first original numeric literal” of 46131 for total bytes in Figure 2 is represented by “a first substitute numeric literal” of 0.22423 for total bytes in Figure 3, and by “a third substitute numeric literal” of -0.98584 for Dev_total_Bytes in Figure 3. Bhatia et al. , then, can be construed to disclose “generating a first substitute numeric literal and a third substitute literal that together represent the first original numeric literal” because an additional “third substitute numeric literal” of -0.98584 for Dev_total_Bytes is derived as a new parameter in Figure 3 representing total bytes of Figure 2, and this “third substitute numeric literal” of -0.98584 for Dev_total_Bytes is subsequently provided in “a sequence of lexical tokens that represent the log entry” to generate by a machine learning model “an inference that characterizes the log entry” as being anomalous or normal. Concerning independent claims 1 and 16, even if Bhatia et al. does not disclose the limitation of generating a first substitute numeric literal “and a third substitute numeric literal that together” represent the first original numeric literal, this is taught by Moore et al. Generally, Moore et al. teaches a similar method and computer program product for analysis of log data using machine learning to determine if a log message is anomalous. (Abstract) However, Moore et al. provides a log entry that includes a plurality of numeric literals on each line of a log entry. That is, Table 2 can be construed to disclose “a log entry” with “a first original numeric literal” of 00000000 c0ab9bc0 00000286 on Line 1 and “a second original numeric literal” of 0x98/0xb0 on Line 2, but “a first original numeric literal” is a ‘compound’ numeric literal of three numeric literals, i.e. , 00000000, c0ab9bc0, and 00000286. Consequently, Moore et al. generates “a first substitute numeric literal” and “a third substitute numeric literal” that include three substitute numeric literals of a number count of 3, an average of 1,077,490,882, and a standard deviation of 1,833,268,393 in Table 3, all of which represent “a first original numeric literal” of 00000000 c0ab9bc0 00000286 in Table 2. Here, Moore et al. provides a way to represent ‘compound’ original numeric literals by a plurality of substitute numeric literals. Compare Specification, ¶[0020] - ¶[0021]: Figure 1, which describes representing compound original lexical tokens so that original numeric lexical token O2 is represented by substitute numeric lexical token S2A and substitute numeric lexical token S2B, or original numeric lexical token O4 is represented by substitute numeric lexical token S4A and substitute numeric lexical token S4B. An objective is to perform anomaly detection to intelligently identify unusual-looking log messages. (¶[0004]) It would have been obvious to one having ordinary skill in the art to generate a third substitute numeric literal that together with a first substitute numeric literal represent a first original numeric literal for an original ‘compound’ numeric literal as taught by Moore et al. to detect an anomaly in an event log of Bhatia et al. for a purpose of performing anomaly detection to intelligently identify unusual-looking log messages. Concerning claims 3 and 23, Moore et al . teaches “the first substitute numeric literal” of ‘1,077,490,882’, and “the third substitute numeric literal” of ‘1,866,268,393’ (¶[0044] - ¶]0046]: Tables 2 and 3); both ‘1,077,490,882’ and‘1,866,268,393’ are ten digit numeric literals that have “identical numeric ranges” between 1,000,000,000 and 2,000,000,000. Concerning claim 4, Moore et al . teaches “the first substitute numeric literal” of ‘1,077,490,882’ is calculated as an average, and “the third substitute numeric literal” of ‘1,866,268,393’ is calculated as a standard deviation (¶[0044] - ¶]0046]: Tables 2 and 3); here, calculating an average and calculating a standard deviation are “distinct respective logics”; that is, an average and a standard deviation are calculated by different algorithms. Concerning claim 5, Moore et al . teaches at least one embodiment of “only one of the first original numeric literal and the third substitute numeric literal exceeds one” in Tables 2 and 3; Line 5 includes “the first original numeric literal of 5 which exceeds one in Table 2, but “the third substitute numeric literal” of standard deviation is ‘0’ on Line 5 of Table 3 (¶[0044] - ¶]0046]: Tables 2 and 3). Similarly, Bhatia et al. discloses an original numeric literal of 46131 for total bytes that exceeds one, but a substitute numeric literal of 0.22423 for total bytes that does not exceed one. Concerning claims 6 and 18, Moore et al . teaches “wherein the log entry contains a third original numeric literal that contains a concatenation of the first original numeric literal and a fourth original numeric literal” for an original numeric literal of ‘00000000 c0ab9bc0 00000286’ in Line 1 of Table 2 (¶[0044] - ¶]0046]: Tables 2 and 3); here, ‘00000000 c0ab9bc0 00000286’ can be construed as “a third original numeric literal that contains a concatenation of the first original numeric literal” of ‘00000000’ and “a fourth original numeric literal” of ‘00000286’. Concerning claim 7, Moore et al . teaches an original numeric lexical token of ‘00000000’and ‘00000286’ in Line 1 of Table 2 (¶[0044] - ¶]0046]: Tables 2 and 3); this lexical token is a concatenation of sixteen individual lexical ‘tokens’; here, sixteen tokens is “at least six numeric lexical tokens”; broadly, a given line of an original message could include an arbitrary number of numeric lexical tokens including more than six numeric lexical tokens. Concerning claims 13 and 20, Bhatia et al. discloses “more numeric lexical tokens than non-numeric lexical tokens” because all of the sequence of lexical tokens that comprise the substitute numeric literals of Figure 3 are numeric values, and none of the lexical tokens are non-numeric lexical tokens in Figure 3. Similarly, Moore et al . teaches “wherein the sequence of lexical tokens contains more numeric lexical tokens than non-numeric lexical tokens” because all of the tokens are numeric in Table 3; here, “the sequence of lexical tokens” is defined by the independent claims to include “the first substitute literal and the second substitute literal”, and “the sequence” is all of the substitute literals in Lines 1 to 5 of Table 3; given that there are no non-numeric tokens in Table 3, then there are “more numeric lexical tokens than non-numeric lexical tokens.” Concerning claim 14, Bhatia et al. discloses “wherein the first original numeric literal” of 46131 for total bytes and “the first substitute numeric literal” of 0.22423 for total bytes “have distinct respective numeric precisions” because 46131 has no decimal point numeric precision in Figure 2 but 0.22423 has a five digit decimal point precision in Figure 3. Similarly, Moore et al . teaches “the first original numeric literal” on Line 3 of Table 2 with a value ‘0.96’ having a numeric precision to two decimal places, but a corresponding “the first substitute numeric literal” of Line 3 of Table 3 can be ‘1’ or ‘0’ that does not include any decimal place precision. (¶[0044] - ¶]0046]: Tables 2 and 3) 07-22-aia AIA Claim s 10 to 11, 19, and 22 are rejected under 35 U.S.C. 103 as being unpatentable over Bhatia et al. (U.S. Patent Publication 2021/0026722) in view of Moore et al. (U.S. Patent Publication 2021/0174253) as applied to claim s 1 and 16 above, and further in view of Pajek et al . (U.S. Patent Publication 2021/0286947) . Concerning claims 10 and 19, Bhatia et al. can be construed to disclose the limitation of “a count of lexical tokens in the sequence of lexical tokens depends on a count of values in the log entry” because there are at least as many lexical tokens representing substitute lexical tokens in Figure 3 as compared to a count of original lexical tokens in a log entry of Figure 2. That is, Bhatia et al. discloses that there are 20 or more features representing attributes of log data and 15 parameters are illustrated for Log 0 in Figure 2. (¶[0054] and ¶[0064]: Figure 2) However, there are at least 25 features that include at least some of the original 15 parameters but include 10 new features in Figure 3. Consequently, Bhatia et al. discloses “a count of lexical tokens in the sequence of lexical tokens” of Figure 3 “depends on a count of values in the log entry” of Figure 2 because there are more parameters, i.e. , 25 features in Figure 3 as compared to 15 features in Figure 2. Concerning claims 10 and 19, Bhatia et al. does not expressly disclose “the inference is a fixed-sized encoding that represents the log entry”. Still, it is known for machine learning to perform encoding as fixed-size vectors in order that features can be uniformly represented in an embedding space, so that this fixed-size encoding may be construed as inherent with inferencing by machine learning of Bhatia et al. Specifically, Pajak teaches that tokens are represented as embeddings of fixed length by passing text 74 through a word embedding model. Pre-processing circuit 64 represents each of the numerical values in array 76 as a respective fixed length vector 86. (¶[0051] - ¶[0057]: Figure 5) Neural network 114 is trained to determine whether a value for a parameter is normal or abnormal. (¶[0088] and ¶[0092]: Figure 8) Here, “the inference . . . that represents the log entry” is whether a value for a parameter is normal or abnormal. Pajak teaches text processing in machine learning with pre-processing of text data for inputting into a trained model including receiving a set of text data that includes numerical information, wherein a first subset of the plurality of tokens comprises tokens that do not comprise numerical information and a second subset of the plurality of tokens comprises tokens that comprise respective numerical information, and each token in the second subset is assigned a respective numerical vector in dependence on the numerical information in the token. (Abstract) Pre-processing circuitry 64 assigns a value of that number as the numerical value for the token for each token that comprises a number. Here, Figure 5 illustrates that a numerical value of ‘22’ is represented as a fixed length vector of ‘2.2’ and ‘1’ and a numerical value of ‘107’ is represented as a fixed length vector of ‘1.07’ and ‘2’. An objective is to deal with numbers when embedding for numbers that are not frequent enough to have a learned embedding. (¶[0018] - ¶[0020]) It would have been obvious to one having ordinary skill in the art to encode a log entry of Bhatia et al. with fixed size encodings as taught by Pajak for a purpose of representing numbers for machine learning embeddings that are not frequent enough to have a learned embedding. Concerning claims 11 and 22, Moore et al . teaches machine-generated logs that include text data describing and/or relating to events that occur during processing by the computing system. A machine-generated log may include a process tag identifying a type of process/event that occurred. (¶[0023]) A log message may may contain a prompt indicating a machine name. (¶[0038]) Textual tokens clustered in a machine-readable log include ‘read’, ‘process’, ‘call’, ‘create’, ‘send’, etc . (¶[0043]: Table 1) Broadly, computer program processes to ‘read’, ‘call’, ‘create’, and ‘send’ are “an execution of a database statement” (“wherein the log entry represents an execution of a database statement”). Similarly, a prompt in machine learning represents “an execution of a database statement.” Allowable Subject Matter 12-151-07 AIA 07-97 12-51-07 Claim s 12 and 24 are allowed. 12-151-08 AIA 07-43 12-51-08 Claim s 8 to 9, 15, and 21 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. Response to Arguments Applicants’ arguments filed 18 May 2026 have been considered but are moot in view of new grounds of rejection as necessitated by amendment. Applicants provide amendments to independent claims 1 and 16 to include new limitations of a log entry “from a plurality of log entries in a log” and “a third substitute numeric literal that together” with a first substitute numeric literal represent a first original numeric literal. Then Applicants present arguments traversing the rejection of these independent claims as being anticipated under 35 U.S.C. §102(a)(1) by Moore et al . (U.S. Patent Publication 2021/0174253). Applicants consider prior art considered during a telephone interview of Bhatia et al. (U.S. Patent Publication 2021/0026722). However, Applicants argue that Bhatia et al. only represents a first original numeric literal of ‘total bytes’ by a first substitute numeric literal, but not “a third substitute numeric literal”. Additionally, Applicants argue that limitations of dependent claim 10 are not rendered obvious in view of Pajak (U.S. Patent Publication 2021/0286947). Here, Applicants argue that Pajak teaches a word embedding model, but a word is not a log entry nor a sequence of lexical tokens. Applicants argue that an extended embedding vector does not represent multiple numbers, i.e. , numeric literals. Similarly, Applicants present an argument against dependent claim 11 that a log message containing a prompt is taught by Moore et al ., but this is a shell command and not a database statement. Moreover, Applicants amend independent method claim 12 to place it into independent form and provide a corresponding computer program product with new independent claim 24, and add a new dependent claim 23. Applicants’ amendments necessitate new grounds of rejection as directed to independent claims 1 and 16 now being obvious under 35 U.S.C. §103 over Bhatia et al. (U.S. Patent Publication 2021/0026722) in view of Moore et al. (U.S. Patent Publication 2021/0174253). Generally, Bhatia et al. addresses the new limitation that specifies that the first and second original numeric literals are from a log entry of a plurality of log entries disclosed by that reference at ¶[0064]: Figure 2. Here, Bhatia et al. clearly discloses that the first and second original numeric literals are obtained from a given Log 0 of a plurality of log entries of Log 0 to Log 4. Then Bhatia et al. discloses first and second substitute numeric literals correspondingly representing the first and second original numeric literals. Bhatia et al. does not clearly disclose a third substitute numeric literal that together with the first substitute numeric literal represent the first original numeric literal. However, Bhatia et al. does disclose that there are more substitute numeric literals in Figure 3 than there are original numeric literals in Figure 2, so that it may be inferred that at least some of these additional substitute numeric literals of Figure 3 are derived from the original numeric literals of Figure 2. Bhatia et al. , then, could be construed to disclose this “third substitute numeric literal” which “together” with “a first substitute numeric literal” represent “a first original numeric literal” because at least some of these additional substitute numeric literals, e.g. , Dev_total_Bytes can be inferred to be derived from the original numeric literal of ‘total bytes’. That is, Figure 3’s ‘Total_bytes’ and ‘Dev_total_bytes’ could be construed as “a first substitute numeric literal” and “a third substitute numeric literal” that together represent ‘total bytes’ of Figure 2. However even if this limitation of “a third substitute numeric literal” to represent “a first original numeric literal” is not disclosed by Bhatia et al. , this is taught by Moore et al. Specifically, Table 3 of Moore et al. teaches that there are three substitute numeric literals of ‘number count’, ‘average’, and ‘standard deviation’ that together represent a first original numeric literal of a line of code of a log entry of Table 2. Here, each line of the log entry can be construed to provide ‘an original numeric literal’, but there is more than one numerical literal on each line of the code. Specifically, a line of code can be understood to include a ‘compound’ original numeric literal, e.g. , Line 1 of Table 2 includes an original numeric literal that is a compound of ‘00000000’, ‘c0ab9bc0’, and ‘00000286’ so that this renders it feasible to represent a ‘compound’ “first original numeric literal” by three “substitute numeric literals” of ‘number count’, ‘average’, and ‘standard deviation’ in Table 3. Similarly, Applicants’ Specification , ¶[0020] - ¶[0021]: Figure 1, describes how compound original lexical tokens can be represented by decomposing the original lexical token that can be viewed as a concatenation of smaller lexical tokens so that there are a plurality of substitute lexical tokens for one original lexical token, e.g. , original lexical token O2 is represented by substitute lexical tokens S2A and S2B and original lexical token O4 is represented by substitute lexical tokens S4A and S4B. Bhatia et al. and Moore et al. are combinable under a rationale of KSR International Co. v. Teleflex Inc. (KSR), 550 U.S. 398, 82 USPQ2d 1385 (2007): (A) Combining prior art elements according to known methods to yield predictable results. It would yield a predictable result to provide a known method of representing a compound original numeric literal by a plurality of substitute numerical literals as taught by Moore et al. for a given original numeric literal in a log entry of a plurality of log entries of Bhatia et al. Applicants arguments are not persuasive as directed against Pajak . Mainly, Applicants are arguing the specifics of the two references individually without properly accounting for what the combination suggests to one having ordinary skill in the art. One cannot show nonobviousness by attacking references individually where the rejections are based on combinations of references. See In re Keller , 642 F.2d 413, 208 USPQ 871 (CCPA 1981); In re Merck & Co., 800 F.2d 1091, 231 USPQ 375 (Fed. Cir. 1986). Generally, it is known in the prior art of machine learning to represent data as fixed length vectors so that the data can be readily compared in a common format. Arguably, fixed length vectors are an implicit characteristic of data representations in machine learning of Bhatia et al. and Moore et al. , but this is expressly taught by Pajak . Here, Pajak is maintained to teach fixed length vector representations of data in machine learning as a general principle without any limitation on what specific type of data is being represented, e.g. , numerical or alphabetical. Moreover, Pajak expressly teaches representing numerical values, e.g. , ‘107’ as fixed length vectors, so that word embeddings are not restricted to ‘words’, nor should a general teaching of representing data as fixed length vectors exclude numeric literals. Consequently, Applicants’ arguments are not persuasive as directed against Pajak . Similarly, Applicants’ arguments are not persuasive that “wherein the log entry represents an execution of a database statement” is omitted by Moore et al. Applicants argue that ¶[0023] of Moore et al. states that a log message may contain a prompt indicating a machine name. However, ¶[0023] of Moore et al. does not actually state anything about a prompt. Instead, ¶[0038] of Moore et al. states that a log message may contain a prompt, but this is only one embodiment taught by Moore et al. , and Applicants do not explain why execution of a prompt cannot be construed as “an execution of a database statement.” A prompt is conventionally a natural language statement provided by a human user and is not a shell command. Nor is ‘a shell prompt’ disclosed by Moore et al. as argued by Applicants. The rejection is maintained to reasonably construe execution of ‘read’, ‘call’, ‘create’, and ‘send’ as “execution of a database statement.” Independent claims 12 and 24 are indicated as allowable. Claims 8, 9, 15, and 21 are indicated as allowable if incorporated into the independent claims. Applicants’ amendments necessitate new grounds of rejection. All of the new grounds of rejection are necessitated by amendment. Accordingly, this rejection is properly FINAL. Conclusion 07-96 The prior art made of record and not relied upon is considered pertinent to Applicants’ disclosure. Wang et al., Adelstein et al., and Harutyunyan et al. disclose related prior art. 07-40 AIA Applicants’ amendment necessitated the new grounds of rejection presented in this Office Action. Accordingly, THIS ACTION IS MADE FINAL . See MPEP §706.07(a). Applicants are 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 MARTIN LERNER whose telephone number is (571) 272-7608. The examiner can normally be reached Monday-Thursday 8:30 AM-6:00 PM. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Richemond Dorvil can be reached at (571) 272-7602. 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. /MARTIN LERNER/Primary Examiner Art Unit 2658 May 28, 2026 Application/Control Number: 18/219,763 Page 2 Art Unit: 2658 Application/Control Number: 18/219,763 Page 3 Art Unit: 2658 Application/Control Number: 18/219,763 Page 4 Art Unit: 2658 Application/Control Number: 18/219,763 Page 5 Art Unit: 2658 Application/Control Number: 18/219,763 Page 6 Art Unit: 2658 Application/Control Number: 18/219,763 Page 7 Art Unit: 2658 Application/Control Number: 18/219,763 Page 8 Art Unit: 2658 Application/Control Number: 18/219,763 Page 9 Art Unit: 2658 Application/Control Number: 18/219,763 Page 10 Art Unit: 2658 Application/Control Number: 18/219,763 Page 11 Art Unit: 2658 Application/Control Number: 18/219,763 Page 12 Art Unit: 2658 Application/Control Number: 18/219,763 Page 13 Art Unit: 2658 Application/Control Number: 18/219,763 Page 14 Art Unit: 2658 Application/Control Number: 18/219,763 Page 15 Art Unit: 2658 Application/Control Number: 18/219,763 Page 16 Art Unit: 2658 Application/Control Number: 18/219,763 Page 17 Art Unit: 2658
Read full office action

Prosecution Timeline

Show 11 earlier events
Jan 26, 2026
Request for Continued Examination
Jan 30, 2026
Response after Non-Final Action
Mar 17, 2026
Non-Final Rejection mailed — §103
Apr 16, 2026
Examiner Interview Summary
Apr 16, 2026
Applicant Interview (Telephonic)
May 18, 2026
Response Filed
Jun 02, 2026
Final Rejection mailed — §103
Jun 30, 2026
Response after Non-Final Action

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12681964
BLOCKWISE CONTROLLED DECODING OF NATURAL LANGUAGE (NL) BASED OUTPUT GENERATED USING A LARGE LANGUAGE MODEL (LLM) TO REDUCE LATENCY IN RENDERING THEREOF
2y 11m to grant Granted Jul 14, 2026
Patent 12675634
Methods and Systems for Graphically Organizing the Meanings of Words
2y 5m to grant Granted Jul 07, 2026
Patent 12670322
SPECIALIZED TOKEN PREDICTION BY A LARGE LANGUAGE MODEL TO PROMPT EXTERNAL INTERVENTION
2y 5m to grant Granted Jun 30, 2026
Patent 12664367
Training a Tokenizer Using Altered Text Data
2y 9m to grant Granted Jun 23, 2026
Patent 12657387
CLASSIFICATION OF COMPLIANCE TEXT DATA WITH FEW SHOT NATURAL LANGUAGE MODEL
3y 10m to grant Granted Jun 16, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

4-5
Expected OA Rounds
78%
Grant Probability
91%
With Interview (+13.4%)
2y 11m (~0m remaining)
Median Time to Grant
High
PTA Risk
Based on 990 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month