DETAILED ACTION
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 11/20/2025 has been entered.
Summary and Status of Claims
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This Office Action is in response to Applicant’s Request for Continued Examination filed 12/22/2025.
Claim 7 is cancelled.
Claims 1-6 and 9-20 are pending.
Claims 1-6 and 9-20 are rejected under 35 U.S.C. 101.
Claims 1, 14-17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al (US Patent 9,619,450)1.
Claims 2-4, 18, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of further in view of Alfonseca et al (US Patent 9,619,450), Goyal et al. (US Patent Pub 2018/0285326).
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al (US Patent 9,619,450), further in view of Herbrich et al. (US Patent Pub 2011/0231405).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al (US Patent 9,619,450), further in view of Elluru et al. (US Patent Pub 2023/0205774).
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al (US Patent 9,619,450), further in view of Ippolito et al. (“Comparison of Diverse Decoding Methods from Conditional Language Models, 6/14/2019).
Claims 10 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al (US Patent 9,619,450), further in view of Salaka et al. (US Patent 11,269,898).
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al (US Patent 9,619,450), further in view of Beeman et al. (US Patent Pub 2009/0319927).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al (US Patent 9,619,450), further in view of Katta et al. (US Patent Pub 2024/0220082).
The text of those sections of Title 35, U.S. Code not included in this action can be found in a prior Office action.
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 1-6 and 9-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more. Determining whether claims are statutory under 35 U.S.C. 101 involves a two-step analysis. Step 1 requires a determination of whether the claims are directed to the statutory categories of invention. Step 2 requires a determination of whether the claims are directed to a judicial exception without significantly more. Step 2 is divided into two prongs, with the first prong having a part 1 and part 2. See MPEP 2106.
Claim 1
Pursuant to Step 2A, part 1, claims are analyzed to determine whether they are directed to an abstract idea. Pursuant to MPEP 2106, claims are deemed to be directed to an abstract idea if, under their broadest reasonable interpretation, they fall within one of the enumerated categories of (a) mathematical concepts, (b) certain methods of organizing human activity, and (c) mental processes. Under the broadest reasonable interpretation, the terms of the claim are presumed to have their plain meaning consistent with the specification as it would be interpreted by one of ordinary skill in the art. See MPEP 2111.
Claim 1 recites the limitations of: (1) receiving an indication of a single subject line from a cloud client of a content generation service, the single subject line comprising a first entity name (2) generating a prompt for a hybrid machine learning model by interleaving soft tokens and hard tokens into an interleaved prompt, (3) generating, by inputting the interleaved prompt into the hybrid machine learning model of the content generation service, a plurality of candidate subject lines associated with the single subject line, wherein the hybrid machine learning model is trained using a dataset of annotated subject lines, and (4) wherein generating the plurality of candidate subject lines comprises: determining that the annotated subject line in the dataset includes a second entity name based at least in part on a set of markers in the annotated subject line that surround the second entity name, (5) adding at least a portion of the annotated subject line comprising the second entity name to a candidate subject line of the plurality of candidate subject lines, (6) replacing the second entity name in the candidate subject line with the first entity name extracted from the single subject line, (7) calculating at least one similarity metric between the single subject line and each candidate subject line of the plurality of candidate subject lines, (8) selecting, by the content generation service, a quantity of candidate subject lines from the plurality of candidate subject lines based at least in part on filtering the plurality of candidate subject lines according to at least one similarity metric associated with each candidate subject line, (9) causing the quantity of candidate subject lines to be displayed at the cloud client, (10) receiving, from the cloud client, feedback associated with the quantity of candidate subject lines and a selection of at least one candidate subject line displayed at the cloud client.
Courts consider a mental process if it “can be performed in the human mind, or by a human using a pen and paper.” The mental process grouping covers concepts performed in the human mind, including observation, evaluation, judgment, and opinion. MPEP 2016(a)(2)(III). Limitations can also be deemed insignificant extra-solution activity (IESA). The term "extra-solution activity" can be understood as activities incidental to the primary process or product that are merely a nominal or tangential addition to the claim. Extra-solution activity includes both pre-solution and post-solution activity. An example of pre-solution activity is a step of gathering data for use in a claimed process, e.g., a step of obtaining information about credit card transactions, which is recited as part of a claimed process of analyzing and manipulating the gathered information by a series of steps in order to detect whether the transactions were fraudulent. An example of post-solution activity is an element that is not integrated into the claim as a whole, e.g., a printer that is used to output a report of fraudulent transactions, which is recited in a claim to a computer programmed to analyze and manipulate information about credit card transactions in order to detect whether the transactions were fraudulent. MPEP 2106.05(g).
Limitations (1) and (9) amount to mere data gathering and data output, which are IESA. Limitation (2) is directed to a generating a prompt for a hybrid machine learning model by interleaving soft and hard tokens. However, the limitation does not recite specifics of how the interleaving is performed and what constitutes a “soft token” and a “hard token.” The specification does not define these terms. Therefore, the broadest reasonable interpretation of “interleaving” encompasses a mental process where “soft tokens” and “hard tokens” can be combined in any manner. Limitation (3) is directed to generating strings associated with a single subject line using a model to calculate similarity where the interleaved prompt is input into the hybrid machine learning model. However, the limitation does not recite specifics of how the model is performed that places meaningful limits on it. Therefore, it amounts to mere steps of applying an abstract idea, which in this case is the mental step of observing a single subject line, evaluating similarity values with other strings, and determining a set of candidate subject lines. Limitations (4) through (7) are directed to the mental step of generating candidate subject lines. The first step includes a determination step requiring acts of evaluation of an annotated subject line to identify a second entity name. The second step requires a step of using part of the annotated subject line in a candidate subject line. The third step requires replacing the second entity name with a first entity name that was previously received. Lastly, limitation (7) requires a mathematical step of calculating a similarity metric between the single subject line and each candidate subject line. These mathematical calculations can be performed in the mind with the through observation and evaluation. Limitation (8) is directed to a step of selecting candidate subject lines based on similarity metrics, which is a mental step where a person can evaluate respective similarity metrics and determine which strings to select. Limitation (10) is directed to receiving data, which is IESA.
The recited content generation service and client is/are recited at a high level of generality, i.e., as a generic components performing generic computer functions.
For at least these reasons, claim 1 is directed to an abstract idea categorized as a mental process.
Pursuant to Step 2A, part 2, claims are analyzed to determine whether the claim as a whole integrates the recited judicial exception into a practical application of the exception. This evaluation is performed by (1) identifying whether there are any additional elements recited in the claim beyond the judicial exception, and (2) evaluating those additional elements individually and in combination to determine whether the claim as a whole integrates the exception into a practical application. See MPEP 2106.04(d). One way to determine integration into a practical application is when the claimed invention improves the functioning of a computer or improves another technology or technical field. To evaluate an improvement to a computer or technical field, the specification must set forth an improvement in technology and the claim itself must reflect the disclosed improvement. See MPEP 2106.04(d)(1).
In this case, as explained above, claim 1 merely recites an abstract idea categorized under mental processes. As discussed above, limitations (1), (9), and (10) are directed to IESA and cannot integrate the abstract idea into a practical application. Limitations (2) through (8) are mental steps that can be practically performed by a person with the aid of a computer as a tool. Generating an interleaved prompt can be done in any manner by a person to combine soft and hard tokens. Use of the model amounts to nothing more than mere instructions to implement an abstract idea on a generic computer. See MPEP 2106.05(f). MPEP 2106.05(f) provides the following considerations for determining whether a claim simply recites a judicial exception with the words “apply it” (or an equivalent), such as mere instructions to implement an abstract idea on a computer: (1) whether the claim recites only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished; (2) whether the claim invokes computers or other machinery merely as a tool to perform an existing process; and (3) the particularity or generality of the application of the judicial exception. Here, the claim recites no details about a particular model. The model is used to generally apply the abstract idea (i.e., perform the similarity calculations) without placing any limitation on how the model operates. In addition, the limitation would cover every mode of implementing the recited abstract idea using a model. The claim omits any details as to how the model solves a technical problem and instead recites only the idea of a solution or outcome. See MPEP 2106.05(f). In the generation of the candidate subject lines, the determination, adding, replacing, and calculating are recited in a manner that can be easily performed by a person in the mind. They also do not recite specific limitations that demonstrate an asserted improvement of a computer or technology. As an example, a person wanting to create a title or headline would look to other examples of titles or headlines they may find interesting or effective and utilize the pattern or template of these examples with the topic/subject of their own content. While claim 1 recite additional components in the form of a content generation service and a cloud client, these components are recited at a high level of generality, which do not add meaningful limits on the recited abstract idea to integrate it into a practical application by providing an improvement to the functioning of a computer or technology, implementing the abstract idea with a particular machine or manufacture that is integral to the claim, effecting a transformation or reduction of a particular article to a different state or thing, nor applying the abstract idea in some meaningful way beyond linking its use to computer technology. See MPEP 2106.04(d).
For at least these reasons, claim 1 does not integrate the judicial exception into a practical application.
Pursuant to Step 2B, claims are analyzed to determine whether the claim as a whole amounts to significantly more than the recited exception i.e., whether any additional element, or combination of additional elements, adds an inventive concept to the claim. See MPEP 2106.05. In this case, claim 1 does not recite limitations that amount to significantly more than the abstract idea. For much of the reasons set forth above, the limitations do not provide an inventive concept. Limitations (1), (9), and (10) are directed to IESA and limitations (2) through (8) are directed to mental steps practically performed by a person. Even when considered in combination, these additional elements represent mere instructions to implement an abstract idea or other exception on a computer and insignificant extra-solution activity, which do not provide an inventive concept. For at least these reasons, claim 1 is nonstatutory because they are directed to a judicial exception without significantly more.
Claims 2-4
Pursuant to step 2A, part 1, claims 2-4 depends on claim 1 and therefore recites the same abstract idea. Pursuant to step 2A, part 2, claims 2-4 recites the additional limitations of calculating the at least one similarity metric comprises: (1) calculating a semantic similarity score between the single subject line and each candidate subject line based at least in part on using the hybrid machine learning model to perform a token-level comparison between the single subject line and the candidate subject line, wherein the similarity metrics comprise the semantic similarity score, (2) calculating a surface form dissimilarity score between the single subject line and each candidate subject line based at least in part on using the hybrid machine learning model to perform a character-level comparison between the single subject line and the candidate subject line, (3) calculating a length consistency score between the single subject line and each candidate subject line. Limitation (1) is directed to performing semantic similarity calculations, which is a mental step or a mathematical calculation, depending on the interpretation of what a semantic similarity score is. These steps are performed by a machine learning model to perform token-level comparison. The limitation amounts to mere steps to “apply it” because it recites using the model without placing any limitations on how the token level comparison is performed. Limitation (2) similarly does not integrate the abstract idea for the same reasons. Limitation (3) is directed to a mathematical calculation without any meaningful limits on how the calculating is performed. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, for much of the same reasons, the limitations do not amount to significantly more than the abstract idea. Accordingly, these additional limitations do not provide an inventive concept. For at least these reasons, claims 2-4 is directed to a judicial exception without significantly more.
Claims 5
Pursuant to step 2A, part 1, claim 5 depends on claim 1 and therefore recites the same abstract idea. Pursuant to step 2A, part 2, claim 5 recites the additional limitations of (1) assigning weights to the at least one similarity metric between the single subject line and the plurality of candidate subject lines based at least in part on using the hybrid machine learning model to calculate a rank correlation between an annotated A/B dataset and the at least one similarity metric. The limitation is directed to assigning weights (i.e., values) to the similarity metrics using a machine learning model. The limitation amounts to steps of “applying” the abstract idea using the model without reciting specific steps of how the rank correlation is calculated or how weights are assigned. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, for much of the same reasons, the limitations do not amount to significantly more than the abstract idea. Accordingly, these additional limitations do not provide an inventive concept. For at least these reasons, claim 5 is directed to a judicial exception without significantly more.
Claims 6
Pursuant to step 2A, part 1, claim 6 depends on claim 1 and therefore recites the same abstract idea. Pursuant to step 2A, part 2, claim 6 recites the additional limitations of (1) wherein selecting the quantity of candidate subject lines comprises: ranking the plurality of candidate subject lines according to a soft majority voting scheme. The limitation is directed to ordering/ranking of strings based on a voting scheme but omits the specifics of how the voting scheme is implemented that would put meaningful limits on the abstract idea. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, for much of the same reasons, the limitations do not amount to significantly more than the abstract idea. Accordingly, these additional limitations do not provide an inventive concept. For at least these reasons, claim 6 is directed to a judicial exception without significantly more.
Claims 9-11
Pursuant to step 2A, part 1, claims 9-11 depends on claim 1 and therefore recites the same abstract idea. Pursuant to step 2A, part 2, claims 9-11 recites the additional limitations of (1) sampling the single subject line and the plurality of candidate subject lines using a decoding algorithm and a temperature control algorithm, (2) determining respective predicted engagement rates for the quantity of candidate subject lines based at least in part on using a performance testing service to evaluate the quantity of candidate subject lines with respect to historic performance data, (3) wherein causing the plurality of candidate subject lines to be displayed comprises: causing the respective predicted engagement rates to be displayed in association with the quantity of candidate subject lines. Limitation (1) is directed to sampling techniques on the strings, such as decoding and temperature control. These algorithms amount to statistical/mathematical algorithms and the limitations do not recite limitations on how these algorithms are performed that would integrate the abstract idea. Limitation (2) is a mental step of determination based on observing engagement rates, evaluating them, and making a judgment. Recitation of the performance testing service is at a high level of generality and does not provide meaningful limits on the abstract idea. Thus, the limitation can practically be performed by a person. Limitation (3) is simply outputting data in a way that associates the engagement rates with the strings. Data output/presentation is IESA and cannot integrate an abstract idea into a practical application on its own. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, for much of the same reasons, the limitations do not amount to significantly more than the abstract idea. Accordingly, these additional limitations do not provide an inventive concept. For at least these reasons, claim 9-11 is directed to a judicial exception without significantly more.
Claim 12
Pursuant to step 2A, part 1, claim 12 depends on claim 1 and therefore recites the same abstract idea. Pursuant to step 2A, part 2, claim 12 recites the additional limitations of (1) causing display of an alert message based at least in part on determining that one or more words in the single subject line are offensive or inappropriate, wherein the alert message comprises a first option to disregard the alert message and a second option to modify the single subject line. While the limitation recites “causing a display of an alert message” a person with the aid of pen and paper, can write down a note (i.e., alert) that the single subject line contains offensive words and an annotation (i.e., alert message) to ignore or go back and modify the single subject line. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, for much of the same reasons, the limitations do not amount to significantly more than the abstract idea. Accordingly, these additional limitations do not provide an inventive concept. For at least these reasons, claim 12 is directed to a judicial exception without significantly more.
Claim 13
Pursuant to step 2A, part 1, claim 13 depends on claim 1 and therefore recites the same abstract idea. Pursuant to step 2A, part 2, claim 13 recites the additional limitations of (1) causing display of a feedback menu based at least in part on a mouse click event associated with an interactive user interface element and (2) receiving, from the cloud client, a selection of one or more options from the feedback menu. Both limitations (1) and (2) are directed to activities of IESA of mere data gathering, where the data in this case is feedback from the user. While the limitations recite a “feedback menu” that is displayed based on a “mouse click event associated with an interactive user interface element” and “receiving … a selection … from the feedback menu”, these additional elements relate to typical user interface elements such as a “thumbs down” or “thumbs up” icon or an option from a drop down menu, as explained in the specification at para. 0067. Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, for much of the same reasons, the limitations do not amount to significantly more than the abstract idea because the elements for receiving feedback (i.e., gathering the data) are known and conventional. Accordingly, these additional limitations do not provide an inventive concept. For at least these reasons, claim 13 is directed to a judicial exception without significantly more.
Claims 14-16
Pursuant to step 2A, part 1, claims 14-16 depends on claim 1 and therefore recites the same abstract idea. Pursuant to step 2A, part 2, claims 14-16 recites the additional limitations of (1) wherein the dataset of annotated subject lines excludes customer data, (2) wherein selecting the quantity of candidate subject lines comprises: selecting all of the candidate subject lines for display based at least in part on the similarity metrics between the single subject line and the plurality of candidate subject lines, (3) wherein the feedback is received from the cloud client via a user interface or an application programming interface. Limitation (1) merely recites the type of data that is to be manipulated and processed, which is IESA. Limitation (2) recites a step of selecting a type of data, based on similarity, to be manipulated/processed, which is IESA. Limitation (3) is directed to a limitation that merely ties the abstract idea into a field of technology (i.e., cloud computing). Therefore, these additional limitations do not integrate the abstract idea into a practical application. Pursuant to step 2B, for much of the same reasons, the limitations do not amount to significantly more than the abstract idea. Accordingly, these additional limitations do not provide an inventive concept. For at least these reasons, claims 14-16 is directed to a judicial exception without significantly more.
Claims 17-19 recite essentially the same subject matter as claims 1-3, respectively, in the form of an apparatus. Therefore, they are rejected for the same reasons. Claim 17 recites the additional components of a processor and a memory, but they are recited at a high level of generality and do not provide meaningful limits on the abstract idea that would integrate it into a practical application or amount to significantly more.
Claim 20 is essentially the same subject matter as claim 1, in the form of a non-transitory computer-readable medium. Therefore, they are rejected for the same reasons. Claim 20 recites the additional components of a processor and a computer readable medium, but they are recited at a high level of generality and do not provide meaningful limits on the abstract idea that would integrate it into a practical application or amount to significantly more.
Claims 1-6 and 9-20 are therefore not drawn to eligible subject matter as they are directed to an abstract idea without significantly more.
To expedite a complete examination of the instant application, the claims rejected under 35 U.S.C. 101 (nonstatutory) above are further rejected as set forth below in anticipation of applicant amending these claims to overcome the rejection.
Note on Prior Art Rejections
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 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.
Claim Rejections - 35 USC § 103
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 of this title, 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.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 14-17, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca).
In regards to claim 1, McAnallen discloses a method for data processing, comprising:
a. receiving an indication of a single subject line from a cloud client of a content generation service, the single subject line comprising a first entity name (McAnallen at paras. 0031, 0038-0040)2;
b. generating, by inputting the received single subject line into the machine learning model of the content generation service, a plurality of candidate subject lines associated with the single subject line (McAnallen at paras. 0041-43, 0045)3, wherein the hybrid machine learning model is trained using a dataset of annotated subject lines (McAnallen at paras. 0025, 0041)4;
c. calculating at least one similarity metric between the single subject line and each candidate subject line of the plurality of candidate subject lines (McAnallen at paras. 0041-43, 0045)5;
d. selecting, by the content generation service, a quantity of candidate subject lines from the plurality of candidate subject lines based at least in part on filtering the plurality of candidate subject lines according to the at least one similarity metric associated with each candidate subject line (McAnallen at para. 0046)6;
e. causing the quantity of candidate subject lines to be displayed at the cloud client (McAnallen at para. 0046)7; and
f. receiving, from the cloud client, feedback associated with the quantity of candidate subject lines and a selection of at least one candidate subject line displayed at the cloud client. McAnallen at para. 0046.8
McAnallen does not expressly disclose generating a prompt for a hybrid machine learning model by interleaving soft tokens and hard tokens into an interleaved prompt and inputting the interleaved prompt into the hybrid machine learning model of the content generation service. As set forth above, McAnallen does disclose inputting into a machine learning model of a content generation service to generate a plurality of titles (i.e., candidate subject lines). What is not expressly disclosed is the machine learning model is a “hybrid machine learning model” and the prompt is generated by “interleaving soft tokens and hard tokens”.
Liu discloses a method of prompt tuning, which generates prompts for a machine learning model by non-invasively concatenating/adding continuous prompts (i.e., soft tokens) with discrete prompts (i.e., hard tokens). Liu at pg. 8, “5.3 language model prompting”.
McAnallen and Liu are analogous art because they are directed to the same field of endeavor of using machine learning models.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen by adding the features of generating a prompt for a hybrid machine learning model by interleaving soft tokens and hard tokens into an interleaved prompt and inputting the interleaved prompt into the hybrid machine learning model of the content generation service, as disclosed by Liu.
The motivation for doing so would have been because the hybrid prompt is useful and superior to discrete prompts (i.e., hard token) alone. Liu at pg. 8, “5.3 Language model prompting”.
McAnallen in view of Liu does not expressly disclose wherein generating the plurality of candidate subject lines comprises: determining that an annotated subject line in the dataset includes a second entity name based at least in part on a set of markers in the annotated subject line that surround the second entity name, adding at least a portion of the annotated subject line comprising the second entity name to a candidate subject line of the plurality of candidate subject lines, and replacing the second entity name in the candidate subject line with the first entity name extracted from the single subject line.
Alfonseca discloses a system and method for automatic generation of headlines based on an input document (i.e., single subject line). The generated headlines (i.e., titles) are based on learned patterns from a corpus of documents (i.e., dataset of annotated documents). The system processes training data to identify patterns of events and entities. This processed training data is used by the headline generation engine to generate headlines (i.e., trained on a dataset of annotated subject lines). Alfonseca at col. 7, lines 37-45. The processing identifies entities in the documents with labels and include markers that indicate that it is an entity label (i.e., second entity name surrounded by a set of markers). Alfonseca at col. 13, lines 18-45; col. 15, lines 42-54. The recognized pattern from the training data is utilized in a generated title (i.e., adding at least a portion … to the candidate subject line) and the marked entity is replaced with a actual entity of an input document (i.e., replacing the second entity name … with the first entity name extracted form the single subject line). Alfonseca at col. 18, lines 22-30.
McAnallen, Liu, and Alfonseca are analogous art because they are directed to the same field of endeavor of machine learning models and processing text with them.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu by adding the features of wherein generating the plurality of candidate subject lines comprises: determining that an annotated subject line in the dataset includes a second entity name based at least in part on a set of markers in the annotated subject line that surround the second entity name, adding at least a portion of the annotated subject line comprising the second entity name to a candidate subject line of the plurality of candidate subject lines, and replacing the second entity name in the candidate subject line with the first entity name extracted from the single subject line, as disclosed by Alfonseca.
The motivation for doing so would have been to provide the benefit of generating headlines that are not subject to copyright and allow for automated generation of headlines while maintaining associations of learned patterns. Alfonseca at col. 2, lines 40-59.
In regards to claim 14, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, wherein the dataset of annotated subject lines excludes customer data. McAnallen at para. 0026.9
In regards to claim 15, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, wherein the quantity of candidate subject lines comprises all of the candidate subject lines of the plurality of candidate subject lines. McAnallen at para. 0046.10
In regards to claim 16, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, wherein the feedback is received from the cloud client via a user interface or an application programming interface. McAnallen at paras. 0038, 0046.11
In regards to claim 17, McAnallen an apparatus for data processing, comprising:
a. at least one processor (McAnallen at para. 0067);
b. at least one memory coupled with the at least one processor (McAnallen at para. 0068); and
c. instructions stored in the at least one memory and executable by the at least one processor (McAnallen at para. 0068) to cause the apparatus to:
i. receive an indication of a single subject line from a cloud client of a content generation service, the single subject line comprising a first entity name (McAnallen at paras. 0031, 0038-0040)12;
ii. generate, by inputting the single subject line into a machine learning model of the content generation service, a plurality of candidate subject lines associated with the single subject line (McAnallen at paras. 0041-43, 0045)13, wherein the hybrid machine learning model is trained using a dataset of annotated subject lines (McAnallen at paras. 0025, 0041)14;
iii. calculate at least one similarity metric between the single subject line and each candidate subject line of the plurality of candidate subject lines (McAnallen at paras. 0041-43, 0045)15;
iv. select, by the content generation service, a quantity of candidate subject lines from the plurality of candidate subject lines based at least in part on filtering the plurality of candidate subject lines according to the at least one similarity metric associated with each candidate subject line (McAnallen at para. 0046)16;
v. cause the quantity of candidate subject lines to be displayed at the cloud client (McAnallen at para. 0046)17; and
vi. receive, from the cloud client, feedback associated with the quantity of candidate subject lines and a selection of at least one candidate subject line displayed at the cloud client. McAnallen at para. 0046.18
McAnallen does not expressly disclose generating a prompt for a hybrid machine learning model by interleaving soft tokens and hard tokens into an interleaved prompt and inputting the interleaved prompt into the hybrid machine learning model of the content generation service. As set forth above, McAnallen does disclose inputting into a machine learning model of a content generation service to generate a plurality of titles (i.e., candidate subject lines). What is not expressly disclosed is the machine learning model is a “hybrid machine learning model” and the prompt is generated by “interleaving soft tokens and hard tokens”.
Liu discloses a method of prompt tuning, which generates prompts for a machine learning model by non-invasively concatenating/adding continuous prompts (i.e., soft tokens) with discrete prompts (i.e., hard tokens). Liu at pg. 8, “5.3 language model prompting”.
McAnallen and Liu are analogous art because they are directed to the same field of endeavor of using machine learning models.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen by adding the features of generating a prompt for a hybrid machine learning model by interleaving soft tokens and hard tokens into an interleaved prompt and inputting the interleaved prompt into the hybrid machine learning model of the content generation service, as disclosed by Liu.
The motivation for doing so would have been because the hybrid prompt is useful and superior to discrete prompts (i.e., hard token) alone. Liu at pg. 8, “5.3 Language model prompting”.
Claim 20 is essentially the same as claim 1 in the form of a non-transitory computer readable medium (McAnallen at para. 0069). Therefore, it is rejected for the same reasons.
Claims 2-4, 18, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca), further in view of Goyal et al. (US Patent Pub 2018/0285326) (Goyal).
In regards to claims 2, 3, and 4, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, but does not expressly disclose calculating the at least one similarity metric comprises: (claim 2) calculating a semantic similarity score between the single subject line and a candidate subject line based at least in part on using the hybrid machine learning model to perform a token-level comparison between the single subject line and the candidate subject line, (claim 3) calculating a surface form dissimilarity score between the single subject line and a character string based at least in part on using the hybrid machine learning model to perform a character-level comparison between the single subject line and the candidate subject line, and (claim 4) calculating a length consistency score between the single subject line and a candidate subject line.
Goyal discloses a system and method for comparing documents to classify and rank changes between document versions. The system provides various methods for determining similarity, such as, term-similarity score for a similarity between each term of the sentences of the documents (i.e., token level comparison …) (Goyal at para. 0047), a character level distance to quantify a number of changes (i.e., character level comparison …) (Goyal at para. 0116), and calculating a length ratio between the two documents being compared (i.e., claim 4). Goyal at para. 0072.
McAnallen, Liu, Alfonseca, and Goyal are analogous art because they are directed to the same field of endeavor of document and text processing.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu and Alfonseca by adding the features of calculating the at least one similarity metric comprises: (claim 2) calculating a semantic similarity score between the single subject line and a candidate subject line based at least in part on using the hybrid machine learning model to perform a token-level comparison between the single subject line and the candidate subject line, (claim 3) calculating a surface form dissimilarity score between the single subject line and a character string based at least in part on using the hybrid machine learning model to perform a character-level comparison between the single subject line and the candidate subject line, and (claim 4) calculating a length consistency score between the single subject line and a candidate subject line, as disclosed by Goyal.
The motivation for doing so would have been to provide a more flexible and nuanced approach to identifying changes between two documents and also provide a more accurate methodology for identifying those changes. Goyal at para. 0131.
Claims 18 and 19 are essentially the same as claims 2 and 3, respectively, in the form of an apparatus. Therefore, they are rejected for the same reasons.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca), further in view of Herbrich et al. (US Patent Pub 2011/0231405) (Herbrich).
In regards to claim 5, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, but does not expressly disclose further comprising: assigning weights to the at least one similarity metric between the single subject line and the plurality of candidate subject lines based at least in part on using the hybrid machine learning model to calculate a rank correlation between an annotated A/B dataset and the at least one similarity metric. Note, as discussed in the rejection of claim 1, McAnallen discloses a machine learning model is used to calculate similarities.
Herbrich discloses a system and method for comparing sketches of users for collaborative filtering to accurately recommend items to users. Herbrich at abstract. The method includes determining a similarity between sketches. The similarity is calculated using rank rorrelations to determine weight rankings according to their similarity (i.e., calculate a rank correlation …). Herbrich at paras. 0038, 0052.
McAnallen, Liu, Alfonseca, and Herbrich are analogous art because they are directed to the same field of endeavor of text processing.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu and Alfonseca by adding the features of assigning weights to the at least one similarity metric between the single subject line and the plurality of candidate subject lines based at least in part on using the hybrid machine learning model to calculate a rank correlation between an annotated A/B dataset and the at least one similarity metric, as disclosed by Herbrich.
The motivation for doing so would have been to provide accurate recommendations to users based on similarities based on rank correlation. Herbrich at para. 0038.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca), further in view of Elluru et al. (US Patent Pub 2023/0205774) (Elluru).
In regards to claim 6, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, but does not expressly disclose wherein selecting the quantity of candidate subject lines comprises: ranking the plurality of candidate subject lines according to a soft majority voting scheme.
Elluru discloses a system and method for leveraging a confidence classifier for information retrieval based on an input query and returning knowledge documents. Elluru at abstract. The system includes the process of determining similarity between a user query and corresponding documents (i.e., single subject line and candidate subject lines) and utilizing a ranking step where the probabilities of each class of the classifiers are calculated. Based on the output probabilities of each class, the documents (i.e., candidate subject lines) are re-franked. Elluru at paras. 0081-86.
McAnallen, Liu, Alfonseca, and Elluru are analogous art because they are directed to the same field of endeavor of document and text processing.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu and Alfonseca by adding the features of wherein selecting the quantity of candidate subject lines comprises: ranking the plurality of candidate subject lines according to a soft majority voting scheme, as disclosed by Elluru.
The motivation for doing so would have been to enable identification of the best question/answer document. Elluru at para. 0081.
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca), further in view of Ippolito et al. (“Comparison of Diverse Decoding Methods from Conditional Language Models, 6/14/2019) (Ippolito).
In regards to claim 9, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, but does not expressly disclose further comprising: sampling the single subject line and the plurality of candidate subject lines using a decoding algorithm and a temperature control algorithm.
Ippolito discloses a system and method for machine learning models that utilize decoding strategies. Decoding strategies are utilized for sampling and use of a temperature parameter to control the entropy of the distribution before sampling. Ippolito at pg. 2, col. 2.
McAnallen, Liu, Alfonseca, and Ippolito are analogous art because they are directed to the same field of endeavor of text processing using machine learning models.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu and Alfonseca by adding the features of sampling the single subject line and the plurality of candidate subject lines using a decoding algorithm and a temperature control algorithm, as disclosed by Ippolito.
The motivation for doing so would have been because it is one of a plurality of standard decoding methods. Ippolito at pg. 2.
Claims 10 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca), further in view of Salaka et al. (US Patent 11,269,898) (Salaka).
In regards to claim 10, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, but does not expressly disclose further comprising: determining respective predicted engagement rates for the quantity of candidate subject lines based at least in part on using a performance testing service to evaluate the quantity of candidate subject lines with respect to historic performance data.
Salaka discloses a system and method for processing a database query and utilizing a machine learning model. The system provides presenting search results in a ranked order, wherein the ranked order is based on user interaction data with the results and prior prediction value that is based on user engagement with a result item. Salaka at Col. 12, lines 47-67.
McAnallen, Liu, Alfonseca, and Salaka are analogous art because they are directed to the same field of endeavor of text processing.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu and Alfonseca by adding the features of determining respective predicted engagement rates for the quantity of candidate subject lines based at least in part on using a performance testing service to evaluate the quantity of candidate subject lines with respect to historic performance data, as disclosed by Salaka.
The motivation for doing so would have been to show results in a manner that indicates how other users have interacted with the same results. Salaka at col. 12, lines 47-67.
In regards to claim 11, McAnallen in view of Liu and Alfonseca and Salaka discloses the method of claim 10, wherein causing the plurality of candidate subject lines to be displayed comprises: causing the respective predicted engagement rates to be displayed in association with the quantity of candidate subject lines. McAnallen at para. 0046. Salaka at col. 12, lines 47-67.19
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca), further in view of Beeman et al. (US Patent Pub 2009/0319927) (Beeman).
In regards to claim 12, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, but does not expressly disclose further comprising: causing display of an alert message based at least in part on determining that one or more words in the single subject line are offensive or inappropriate, wherein the alert message comprises a first option to disregard the alert message and a second option to modify the single subject line.
Beeman discloses a system and method for processing documents based on a set of rules. Beeman at abstract. The method includes determining whether an input document contains content that is a violation of the rule (i.e., inappropriate words) and sending the user a notification (i.e., alert). In these cases, the system can be configured by the administrator to control whether the user can ignore a violation notification and distribute the document anyway (i.e., disregard the alert message) or blocking the user but allowing the user to edit the document to continue the submission (i.e., second option to modify …). Beeman at para. 0036.
McAnallen, Liu, Alfonseca, and Beeman are analogous art because they are directed to the same field of endeavor of document and text processing.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu and Alfonseca by adding the features of causing display of an alert message based at least in part on determining that one or more words in the single subject line are offensive or inappropriate, wherein the alert message comprises a first option to disregard the alert message and a second option to modify the single subject line, as disclosed by Beeman.
The motivation for doing so would have been to allow a system administrator to control the inputs received from users. Beeman at para. 0036.
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over McAnallen (US Patent Pub 2024/0104055), in view of Liu et al. (“GPT Understands, Too”, 2021) (Liu), further in view of Alfonseca et al. (US Patent 9,619,450) (Alfonseca), further in view of Katta et al. (US Patent Pub 2024/0220082) (Katta).
In regards to claim 13, McAnallen in view of Liu and Alfonseca discloses the method of claim 1, but does not expressly disclose wherein receiving the feedback comprises:
a. causing display of a feedback menu based at least in part on a mouse click event associated with an interactive user interface element; and
b. receiving, from the cloud client, a selection of one or more options from the feedback menu.
Katta discloses a system and method for semantically sorting content, such as documents. Katta at abstract. The system utilizes machine learning methods to calculate similarity, which is used for sorting the documents. Katta at para. 0025. Once the sorted documents are displayed to the user, the user can provide feedback. Feedback can be provided using a GUI thumbs up/thumbs down menu or a drop-down menu. This feedback is used to retrain the model. Katta at para. 0040.
McAnallen, Liu, Alfonseca, and Katta are analogous art because they are directed to the same field of endeavor of document processing and similarity determination.
At the time before the effective filing date of the instant application, it would have been obvious to one of ordinary skill in the art to modify McAnallen in view of Liu and Alfonseca by adding the features of wherein receiving the feedback comprises: causing display of a feedback menu based at least in part on a mouse click event associated with an interactive user interface element and receiving, from the cloud client, a selection of one or more options from the feedback menu, as disclosed by Katta.
The motivation for doing so would have been to allow retraining of the model to reflect user preferences. Katta at para. 0040.
Response to Amendment
Specification
Applicant’s amendment to the specification is acknowledged and accepted.
Response to Arguments
Rejection of Claims 1-7 and 9-20 under 35 U.S.C 101
Claim 7 is cancelled rendering its rejection moot.
Applicant’s arguments in regards to the rejection of claims 1-6 and 9-20 under 35 U.S.C. 101 have been fully considered but they are not persuasive. Applicant alleges the claims are patent eligible because they recite features that integrate the abstract idea into a practical application because they include improvements to the technical field of sensitive data security. Applicant argues that by determining that an annotated subject line includes a second entity name and replacing the second entity name with an extracted first entity name from the single subject line, the recited model produces appealing subject lines and encourages safety. Remarks at 12. Examiner respectfully disagrees.
Applicant points to paras. 0041 and 0048, which correspond to paras. 0042 and 0049 of the originally filed specification. These paragraphs discuss ensuring data sensitivity because the training dataset is devoid of customer data and that producing appealing subject lines and encourages safety (which seems to be for brand consistency), is achieved by replacing marked entities with a placeholder. While there an asserted improvement discussed and described in the specification, the claims do not currently recite limitations that sufficiently demonstrate such improvements. While the claim recites replacing a second entity name with a first entity name, there are no recitations of utilizing some type of generic placeholder. Even if such steps were included, steps of replacing named entities with generic placeholders is known in the art, such as disclosed in Alfonseca discussed in the rejections above. Moreover, the discussion at para. 0049 seems to amount to nothing more than conclusory statement without required details to make the improvement apparent to one of ordinary skill in the art. MPEP 2106.05(a). Furthermore, while a dependent claim recites that the annotated dataset excludes customer data, such a feature in and of itself does not improve the technology because it merely specifies what type of data is used in the claimed process. Accordingly, the claims do not recite limitations that when considered alone or in combination, integrate the abstract idea into a practical application that demonstrates an improvement to a computer or the technology.
For at least these reasons, Examiner asserts claims 1-6 and 9-20, as amended, do not integrate the abstract idea into a practical application. Consequently, the rejection to claims 1-6 and 9-20 under 35 U.S.C. 101 is maintained.
Rejection of claims 1-7 and 9-20 under 35 U.S.C. 103
Claim 7 is cancelled rendering its rejection moot.
Applicant’s arguments in regards to the rejections to claims 1-6 and 9-20 under 35 U.S.C. 103, have been fully considered and they are not persuasive. Additionally, new grounds of rejection set forth above as necessitated by Applicant’s amendments, which add new limitations to the independent claims. Examiner acknowledges that the limitations are similar to those recited in cancelled claim 7, the limitations are not identical. The new grounds of rejection rely on Alfonseca which discloses a system and method for automatically generating a headline for an input document based on learned patterns and identified entities of a processed dataset of documents (i.e., training data).
Applicant alleges McAnallen does not disclose (1) “generating a ‘plurality of candidate subject lines associated with [a] single subject line’” (Remarks at 14) and (2) “calculating a at least one similarity metric between the single subject line and each candidate subject line of the plurality of candidate subject lines.” Remarks at 15. The Examiner respectfully disagrees.
Examiner is required to give claim limitations their broadest reasonable interpretation in light of the specification. However, limitations from the specification are not read into the claims. MPEP 2111.
In regards to limitation (1), McAnallen discloses transmitting a document cluster/group to the system for titles to be generated for each document in the cluster/group. McAnallen at paras. 0025, 0040. Clusters have a topic or topics for by which the documents are grouped. This topic is interpreted as a “subject” of the documents (i.e., line). McAnallen at para. 0039. Titles (i.e., candidate subject lines) are generated for each document in the cluster. These titles are associated with the cluster topic (i.e., associated with the single subject line). The process of selecting the top one or few titles involves calculating a similarity between each generated title embedding (i.e., each candidate subject line) and the average cluster embedding (i.e., single subject line). McAnallen at paras. 0043-45. For at least these reasons, McAnallen discloses limitation (1).
In regards to limitation (2), McAnallen discloses, as Applicant notes, calculating similarity between each of the generated title embeddings (i.e., candidate subject lines) and the average cluster embedding (i.e., single subject line). Based on the explanation above, contrary to Applicant’s allegations, McAnallen discloses limitation (2).
Applicant does not present additional arguments in regards to the remaining limitations. Therefore, claim 1 remains rejected under the new grounds of rejection set forth above as necessitated by Applicant’s amendments. Applicant also does not present arguments in regards to the remaining claims. Therefore, they remain rejected under the new grounds of rejection set forth above as well.
Consequently, the rejection to claims 1-6 and 9-20 under 35 U.S.C. 103 is maintained under the new grounds of rejection set forth above as necessitated by Applicant’s amendments.
Additional Prior Art
Additional relevant prior art are listed on the attached PTO-892 form. Some examples are:
Hu et al. (US Patent Pub 2021/0174025) discloses a system and method for hierarchical entity recognition and semantic modeling for information extraction including entity recognition.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Examiner Michael Le whose telephone number is 571-272-7970 and fax number is 571-273-7970. The examiner can normally be reached Mon-Fri 9:30 AM – 6 PM.
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, Tony Mahmoudi can be reached on 571-272-4078. 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.
/MICHAEL LE/Examiner, Art Unit 2163
/TONY MAHMOUDI/Supervisory Patent Examiner, Art Unit 2163
1 Previously provided on the PTO-892 form mailed 2/13/2025.
2 User creates or transmits (i.e., receiving an indication) a document (i.e., single subject line) to the server. A document can contain entities, topics/subjects (i.e., first entity name). User interacts with the system via a cloud application (i.e., cloud client).
3 The documents are sent to the title generation model (i.e., machine learning model) and titles are generated (i.e., candidate subject lines).
4 The model is trained using labeled documents and/or user generated titles for documents (i.e., a dataset of annotated subject lines).
5 Similarity is calculated for each of the generated titles with reference to a reference embedding generated for the input documents (i.e., calculating at least one similarity metric between single subject line and each of the plurality of candidate subject lines).
6 The generated titles are ranked based on similarity and the top one or few titles are selected as candidate title.
7 The selected titles are output and sent to the user for review and selection.
8 Titles presented to the user for selection also receive feedback relating to the selection for ongoing training of the model.
9 The models are trained on data that does not include user data (i.e., annotated subject lines excludes customer data).
10 The generated titles with the highest similarities are selected for display.
11 User provides feedback via the cloud application interface.
12 User creates or transmits (i.e., receiving an indication) a document (i.e., single subject line) to the server. A document can contain entities, topics/subjects (i.e., first entity name). User interacts with the system via a cloud application (i.e., cloud client).
13 The documents are sent to the title generation model (i.e., machine learning model) and titles are generated (i.e., candidate subject lines).
14 The model is trained using labeled documents and/or user generated titles for documents (i.e., a dataset of annotated subject lines).
15 Similarity is calculated for each of the generated titles with reference to a reference embedding generated for the input documents (i.e., calculating at least one similarity metric between single subject line and each of the plurality of candidate subject lines).
16 The generated titles are ranked based on similarity and the top one or few titles are selected as candidate title.
17 The selected titles are output and sent to the user for review and selection.
18 Titles presented to the user for selection also receive feedback relating to the selection for ongoing training of the model.
19 Ranked prediction value based on user interaction is used for the ranked display of results.