Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
This communication is a Non-Final Office Action in response to Applicant’s application number 18/928,139 received on 10/27/2024.
Claims 1-20 are currently pending and have been examined.
Priority
Applicants claim for the benefit of a prior-filed application under 35 U.S.C. 119 and/or 35 U.S.C. 120 is acknowledged.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 10/27/2024 and 10/23/2025 have been considered by the examiner.
Claim Objections
Claim 9/10/19 are objected to because of the following informalities: the limitations inputting the low-dimensional target user feature into a decoder, to obtain a high-dimensional target user feature and calculating an error between the user feature and the high-dimensional user feature; are duplicated in the claims. Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-patentable subject matter. The claims are directed to an abstract idea without significantly more. The judicial exception is not integrated into a practical application. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception as further set forth in MPEP 2106.
Step 1: The claimed invention is analyzed to determine if it falls outside one of the four statutory categories of invention. See MPEP 2106.03
Claim(s) 1-10 is/are directed to a method (i.e., Process), claim(s) 11-19 is/are directed to a computer device (i.e., Manufacture), and claim(s) 20 is/are directed to a non-transitory computer-readable storage medium (i.e., Manufacture). Therefore, claims 1-20 are directed to patent eligible categories of invention. Accordingly, the claims satisfy Step 1 of the eligibility inquiry.
Step 2A, Prong 1: In prong one of step 2A, the claim(s) is/are analyzed to evaluate whether they recite a judicial exception. See MPEP 2106.04
Independent claims 1, 11, and 20 recite a method, a non-transitory computer-readable recording medium, and a system for collecting data. As drafted, the limitations recited by claims 1, 11, and 20 fall under the “Mental Processes” abstract idea group by setting forth activities that could be performed mentally by a human (including an observation, evaluation, judgment, opinion) (see MPEP § 2106.04(a)(2), subsection III).
Independent claim 1 recites a user identification method performed by a computer device with the following limitations:
obtaining a plurality of users of an application program; (The step for “obtaining a plurality of users of an application program” could be accomplished mentally, such as by human observation, evaluation, judgement, or with the help of pen and paper.);
generating, an account feature and a service feature of a user from the plurality of users, the account feature being based on account use information of the user in the application program, and the service feature being based on a historical payment-related behavior of the user; (The step for “generating an account feature and a service feature” could be accomplished mentally, such as by human observation, evaluation, judgement, or with the help of pen and paper.);
aggregating account features in a plurality of time dimensions to obtain a target account feature; (The step for “aggregating account features” could be accomplished mentally, such as by human observation, evaluation, judgement, or with the help of pen and paper.);
aggregating service features in the plurality of time dimensions to obtain a target service feature; (The step for “aggregating service features” could be accomplished mentally, such as by human observation, evaluation, judgement, or with the help of pen and paper.);
generating an user feature of the user based on the target account feature and the target service feature; (The step for “generating an user feature” could be accomplished mentally, such as by human observation, evaluation, judgement, or with the help of pen and paper.);
and inputting user features of the plurality of users into a classifier to obtain a predicted classification result, the predicted classification result indicating whether the plurality of users are users with a propensity to pay. (But for the recitation of additional elements recited in this claim limitation (underlined), the step to “obtain a predicted classification result” could be accomplished mentally, such as by human observation, evaluation, judgement, or with the help of pen and paper.).
Independent claims 11 and 20 recite a computer device comprising a processor and a memory, and a non-transitory computer-readable storage medium with limitations that are substantially similar to the limitations recited by independent claim 1, therefore the same analysis applies. The additional elements beyond the abstract idea for consideration under Step 2A, Prong 2, and Step 2B recited by the independent claims are:
From claims 1/11: computer device
From claims 1/11/20: classifier
From claim 11: memory
From claims 11/20: processor
From claim 20: non-transitory computer-readable storage medium
Dependent claims 5/15 further narrows the abstract idea and introduce the following additional elements for consideration under said steps: online meeting software.
Dependent claims 9, 10, and 19 further narrows the abstract idea and introduce the following additional elements for consideration under said steps: encoder, and decoder.
Dependent claims 2-4, 6-8, 12-14, and 16-18 further narrow the abstract idea and do not introduce any additional elements for consideration under said steps. In other words, each of the limitations/elements recited in respective dependent claims is/are further part of the abstract ideas as identified by the Examiner for each respective dependent claim (i.e., they are part of the abstract idea recited in each respective claim).
Step 2A, Prong 2: An evaluation is made whether a claim recites any additional element, or combination of additional elements, that integrate the judicial exception into a practical application of the exception. See MPEP 2106.04(d).
Regarding the computing additional elements, namely computer device from claims 1/11, processor from claims 11/20, memory from claim 11, and non-transitory computer-readable storage medium from claim 20, these additional elements have been evaluated but fail to integrate the abstract idea into a practical application because they amount to using generic computing elements or instructions (software) to perform the abstract idea, similar to adding the words “apply it” (or equivalent), which merely serves to link the use of the judicial exception to a particular technological environment (generic computing environment). See MPEP 2106.05(f) and 2106.05(h).
With respect to the limitations from independent claims 1/11/20: “and inputting user features of the plurality of users into a classifier to obtain a predicted classification result”, these limitations have been considered under Step 2A Prong Two, however these limitations invoke computers or other machinery (e.g., machine learning) merely as a tool to perform the abstract idea (e.g., mental processes), which fails to provide a technical improvement or otherwise integrate the abstract idea into a practical application. Use of a computer or other machinery in its ordinary capacity for economic or other tasks or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., mental processes) does not integrate a judicial exception into a practical application. See MPEP 2106.05(f).
With respect to the limitations from dependent claims 5/15: “wherein the application program comprises online meeting software“, these limitations have been evaluated but fail to integrate the abstract idea into a practical application because they amount to using generic computing elements or instructions (e.g., software) to perform the abstract idea, similar to adding the words “apply it” (or equivalent), which merely serves to link the use of the judicial exception to a particular technological environment (generic computing environment). See MPEP 2106.05(f) and 2106.05(h).
With respect to the limitations from dependent claims 9/19: “inputting, for one of the plurality of user features, the user feature into an encoder, to obtain a low-dimensional user feature;”, “inputting the low-dimensional user feature into a decoder, to obtain a high-dimensional user feature;”, “adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition;”, “re-performing the inputting the user feature into an encoder, to obtain a low-dimensional user feature;”, “and inputting the low-dimensional user feature into a decoder, to obtain a high-dimensional user feature”; and from claim 10: “inputting, for one of the plurality of target user features, the target user feature into an encoder, to obtain a low-dimensional target user feature; ”, “inputting the low-dimensional target user feature into a decoder, to obtain a high-dimensional target user feature; ”, “adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition; ”, “re-performing the inputting the target user feature into an encoder, to obtain a low-dimensional target user feature; ”, and “inputting the low-dimensional target user feature into a decoder, to obtain a high-dimensional target user feature.“, these have been considered under Step 2A Prong Two, however these limitations invoke computers or other machinery merely as a tool to perform the abstract idea (e.g., mental processes), which fails to provide a technical improvement or otherwise integrate the abstract idea into a practical application. Use of a computer or other machinery in its ordinary capacity for economic or other tasks or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., mental processes) does not integrate a judicial exception into a F application. See MPEP 2106.05(f).
Accordingly, because the Step 2A Prong One and Prong Two analysis resulted in the conclusion that the claims are directed to an abstract idea, additional analysis under Step 2B of the eligibility inquiry must be conducted in order to determine whether any claim element or combination of elements amount to significantly more than the judicial exception.
Step 2B: The claims are analyzed to determine whether any additional element, or combination of additional elements, is/are sufficient to ensure that the claims amount to significantly more than the judicial exception. This analysis is also termed a search for "inventive concept." See MPEP 2106.05.
Regarding the computing additional elements, namely computer device from claims 1/11, processor from claims 11/20, memory from claim 11, and non-transitory computer-readable storage medium from claim 20, these additional element(s) has/have been evaluated, but fail to add significantly more to the claims because they amount to using generic computing elements (computer hardware) or instructions/software (engine) to perform the abstract idea, similar to adding the words “apply it” (or an equivalent), which merely serves to link the use of the judicial exception to a particular technological environment (network computing environment, the internet, online) and does not amount to significantly more than the abstract idea itself. Applicant’s specification recites the computing additional elements at a high level of generality. Therefore, the additional elements merely describe generic computing elements or computer-executable instructions (software) merely serve to tie the abstract idea to a particular operating environment, which does not add significantly more to the abstract idea. See, e.g., Alice Corp., 134 S. Ct. 2347, 110 USPQ2d 1976; Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015).
With respect to the limitations from independent claims 1/11/20: “and inputting user features of the plurality of users into a classifier to obtain a predicted classification result”; these claim limitations have been considered under Step 2B, however these limitations invoke computers or other machinery (e.g., machine learning) merely as a tool to perform the abstract idea (e.g., mental processes), which fails to provide a technical improvement or otherwise add significantly more to the abstract idea. Use of a computer or other machinery in its ordinary capacity for economic or other tasks or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., mental processes) does not add significantly more to the abstract idea. See MPEP 2106.05(f).
With respect to the limitations from dependent claims 5/15: “wherein the application program comprises online meeting software“, these limitations has/have been evaluated, but fail to add significantly more to the claims because they amount to using generic computing elements (computer hardware) or instructions/software (engine) to perform the abstract idea, similar to adding the words “apply it” (or an equivalent), which merely serves to link the use of the judicial exception to a particular technological environment (network computing environment, the internet, online) and does not amount to significantly more than the abstract idea itself. Applicant’s specification recites the computing additional elements at a high level of generality. Therefore, the additional elements merely describe generic computing elements or computer-executable instructions (e.g., software) merely serve to tie the abstract idea to a particular operating environment, which does not add significantly more to the abstract idea. See, e.g., Alice Corp., 134 S. Ct. 2347, 110 USPQ2d 1976; Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306, 1334, 115 USPQ2d 1681, 1701 (Fed. Cir. 2015).
With respect to the limitations from dependent claims 9/19: “inputting, for one of the plurality of user features, the user feature into an encoder, to obtain a low-dimensional user feature;”, “inputting the low-dimensional user feature into a decoder, to obtain a high-dimensional user feature;”, “adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition;”, “re-performing the inputting the user feature into an encoder, to obtain a low-dimensional user feature;”, “and inputting the low-dimensional user feature into a decoder, to obtain a high-dimensional user feature”; and from claim 10: “inputting, for one of the plurality of target user features, the target user feature into an encoder, to obtain a low-dimensional target user feature; ”, “inputting the low-dimensional target user feature into a decoder, to obtain a high-dimensional target user feature; ”, “adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition; ”, “re-performing the inputting the target user feature into an encoder, to obtain a low-dimensional target user feature; ”, and “inputting the low-dimensional target user feature into a decoder, to obtain a high-dimensional target user feature.“, these claim limitations have been considered under Step 2B, however these limitations invoke computers or other machinery merely as a tool to perform the abstract idea (e.g., mental processes), which fails to provide a technical improvement or otherwise add significantly more to the abstract idea. Use of a computer or other machinery in its ordinary capacity for economic or other tasks or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., mental processes) does not add significantly more to the abstract idea. See MPEP 2106.05(f).
In addition, when taken as an ordered combination, the ordered combination adds nothing that is not already present as when the elements are taken individually. Their collective functions merely provide generic computer implementation. Therefore, when viewed as a whole, these additional claim elements do not provide meaningful limitations to amount to significantly more than the abstract idea itself.
Dependent claims 2-4, 6-8, 12-14, and 16-18 recite the same abstract ideas as the independent claims along with further steps/details falling under the scope of the abstract idea itself and an additional abstract idea along with the same or substantially same generic computing element addressed above under Step 2A Prong Two and Step 2B, which is incorporated herein.
The ordered combination of elements in the claims (including the limitations inherited from the parent claim(s)) add nothing that is not already present as when the elements are taken individually. There is no indication that the combination of elements improves the functioning of a computer or improves any other technology. Their collective functions merely provide generic computer implementation. Accordingly, the subject matter encompassed by the dependent claims fails to amount to significantly more than the abstract idea itself.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-6, 11-16, and 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Swierk et al. (US 11350058 B1, hereinafter “Swierk”), in view of Qiu et al. (US 20220129318 A1, hereinafter “Qiu”).
Regarding claims 1/11/20: Swierk teaches a user identification method ([Column 3, Lines 20-27] A method is needed for tailoring the video capture and video processing methods at each individual computing device participating in the user session to optimize performance of both the multimedia multi-user collaboration application and other concurrently executed applications, based on performance metrics at each computing device, and on statistics describing each user's participation within the current user session.), a computer device comprising a processor and a memory ([Column 10, Lines 3-10] The information handling system may include memory (volatile (e.g., random-access memory, etc.), nonvolatile (read-only memory, flash memory etc.) or any combination thereof), one or more processing resources, such as a central processing unit (CPU), a graphics processing unit (GPU), a vision processing unit (VPU), a Gaussian neural accelerator (GNA), hardware or software control logic, or any combination thereof.), and a non-transitory computer-readable storage medium ([Column 15, Lines 22-25] The disk drive unit 106 and the intelligent collaboration contextual session management system 170 may include a computer-readable medium 172 in which one or more sets of instructions 174 such as software may be embedded.) with the following limitations:
obtaining a plurality of users of an application program; ([Fig. 8] Step 804; [Column 48, Lines 43-45] The values gathered at block 804 may identify a number of participants)
generating, an account feature and a service feature of a user from the plurality of users, the account feature being based on account use information of the user in the application program, ([Column 48, Lines 43-51] The values gathered at block 804 may identify a number of participants, varying presentation displays, percentage of time users spend on mute, percentage of time users actively engage with the multimedia multi-user collaboration application (e.g., using a messaging service incorporated therewithin), identification of one or more users as a host, or identification of one or more users as sharing screens, for example. Examiner notes that one of ordinary skill in the art would reasonably interpret the percentage of time a user spend on mute as an account feature, and the percentage of time users actively engage with the multimedia multi-user collaboration application as a service feature.).
aggregating account features in a plurality of time dimensions to obtain a target account feature; ([Column 34, Lines 21-29] Such meeting metrics may include, for example, a measure of the CPU resources consumed by the multimedia multi-user collaboration application over time. Other example meeting metrics may include a measure of memory resources consumed. Still other example meeting metrics may compare CPU or memory usage by the multimedia multi-user collaboration application 550 to total CPU or memory used by all applications, hardware, or firmware during the training user session.).
and inputting user features of the plurality of users into a classifier to obtain a predicted classification result ([Column 46, Line 9] machine-learning classifier neural network; [Column 46, Lines 31-33] The neural network may include a plurality of layers, including an input layer, one or more hidden layers, and an output layer.; [Column 45, Lines 12-15] measure of the time spent muted, or whether the user is sharing his or her screen, whether a user has the camera on, or whether user's gaze is at the screen. Examiner notes that one of ordinary skill in the art would consider the camera on, the user’s gaze at the screen, and whether the user is sharing his or her screen as a plurality of features that are used as input into the classifier neural network.).
Swierk doesn’t explicitly teach:
and the service feature being based on a historical payment-related behavior of the user;
aggregating service features in the plurality of time dimensions to obtain a target service feature;
generating an user feature of the user based on the target account feature and the target service feature;
the predicted classification result indicating whether the plurality of users are users with a propensity to pay.
Qiu teaches:
and the service feature being based on a historical payment-related behavior of the user; ([0089] The platform user can also be represented by one or more billing accounts. Billing accounts representing the platform user can include data corresponding to a number of different features characterizing billing behavior by the platform user on the platform 110. Examples include the number of billing accounts representing the platform user, the age of each billing account, reported fraudulent activity in relation to paying bills owed for subscribed-to services, a rolling balance for each billing account, and the payment (or non-payment history) of the platform user.);
aggregating service features in the plurality of time dimensions to obtain a target service feature; ([0089] a rolling balance for each billing account, and the payment (or non-payment history) of the platform user);
generating an user feature of the user based on the target account feature and the target service feature; ([0105] the feature engineering module 215 uses data across a plurality of time-steps corresponding to the platform user to determine statistical trends. For example, the feature engineering module 215 can generate a value for a feature representing a trend of the platform user's resource consumption over a period of time. This and other features can then be processed by the scoring-serving model to predict future behavior, i.e., the likelihood of abuse by the platform user given additional resources. In some implementations, the feature engineering module 215 uses data across a plurality of different platform users to generate trends or baselines from which to compare with feature data.);
the predicted classification result indicating whether the plurality of users are users with a propensity to pay. ([0022] The predicted likelihood can be a predicted abuse likelihood, the operations can include maintaining a resource allocation service for allocating additional computing resources of the computing platform in response to a user request and a payment. Generating the score can include generating a predicted revenue likelihood that the computing platform will receive a first request to allocate additional computing resources and a first payment from one of the one or more user accounts within the predetermined future time period, and generating the score as a function comprising the predicted revenue likelihood and the predicted abuse likelihood.; [0077] Features can be categorized according to the type of account from the feature engineering module 215 generates values from, e.g., platform account level features, billing account level features, etc.; [0078] Platform account level features are generally features related to how the platform user accesses the platform 110, and how the platform 110 identifies and authenticates the platform user.; [0079] Platform level account features can also include features characterizing information about a platform account of a platform user.; [0082] Another type of platform account level feature relates to cross-product usage by the platform user of different services or applications hosted by the platform 110.; [0089] The platform user can also be represented by one or more billing accounts. Billing accounts representing the platform user can include data corresponding to a number of different features characterizing billing behavior by the platform user on the platform 110. Examples include the number of billing accounts representing the platform user, the age of each billing account, reported fraudulent activity in relation to paying bills owed for subscribed-to services, a rolling balance for each billing account, and the payment (or non-payment history) of the platform user. Similar to a platform account, billing accounts representing the platform user can also represent when a billing account was created, and how often the platform user is active on the platform 110 in performing billing-related activities.; [0090] billing account features can represent how much of the subscription/purchasing is performed by the platform user versus another user).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine Swierk with Qiu’s feature(s) listed above. One would’ve been motivated to do so in order to successfully aggregate any of the above-described features to represent an aggregated feature across all user accounts representing the platform user (Qiu; [0096]). By incorporating the teachings of Qiu, one would’ve been able to successfully use target features to create user features in order to predict propensity to pay.
Regarding claims 2/12: The combination of Swierk and Qiu teaches all the limitations of claims 1/11 above. Swierk further teaches:
performing, based on values of first-type account features respectively corresponding to the plurality of users, unified encoding on first-type account features having values in a same value range in the plurality of first-type account features, to obtain an encoded first-type account feature; ([Column 48, Lines 29-57] teaches gathering all the inputs for the neural network from the multimedia processing control API and multimedia multi-user collaboration application at block 804, including percentage of time users spend on mute, percentage of time users actively engage with the multimedia multi-user collaboration application (e.g., using a messaging service incorporated therewithin), identification of one or more users as a host, or identification of one or more users as sharing screens.; [Column 6, Line 54-Column 7, Line 3] teaches the processing of multimedia data captured, which may include an encoding process. Examiner notes that one of ordinary skill in the art would reasonably interpret the percentage of time users spend on mute as a first-type account feature, and the percentage representation as the encoding of that feature.);
and performing, based on a type of the second-type account feature, unified encoding on second-type account features belonging to a same category in the plurality of second-type account features, to obtain an encoded second-type account feature; ([Column 48, Lines 29-57] teaches gathering all the inputs for the neural network from the multimedia processing control API and multimedia multi-user collaboration application at block 804, including percentage of time users spend on mute, percentage of time users actively engage with the multimedia multi-user collaboration application (e.g., using a messaging service incorporated therewithin), identification of one or more users as a host, or identification of one or more users as sharing screens.; [Column 6, Line 54-Column 7, Line 3] teaches the processing of multimedia data captured, which may include an encoding process. Examiner notes that one of ordinary skill in the art would reasonably interpret the percentage of time users actively engage with the multimedia multi-user collaboration application as a second-type account feature and the percentage representation as the encoding of that feature.);
the aggregating account features in a plurality of time dimensions to obtain a target account feature comprises: aggregating encoded first-type account features in the plurality of time dimensions, to obtain a target first-type account feature; ([Column 34, Lines 29-32] Yet other example meeting metrics may measure participation of the user during a user session, including, for example, a measure of the time spent muted, or whether the user is sharing his or her screen.);
and aggregating encoded second-type account features in the plurality of time dimensions, to obtain a target second-type account feature; ([Column 50, Lines 18-25] time spent engaging with the multimedia multi-user collaboration application (e.g., time spent actively speaking, time spent sharing screens, time spent directing gaze toward multimedia multi-user collaboration application, or time spent providing input into multimedia multi-user collaboration application GUI) may be given a positive value between zero and positive one.);
and the generating a user feature of the user based on the target account feature and the target service feature comprises generating the user feature of the user based on the target first-type account feature, the target second-type account feature, and the target service feature. ([Column 50, Lines 10-30] the user participation level may represent some combination of the user participation levels described directly above. For example, the user participation levels measured as a percentage of time during the current user session (e.g., time spent muted, time spent directing gaze toward camera, time spent actively speaking or sharing, or time interacting with a GUI for another application) may be represented in decimal form in a scale from negative one to positive one. In such an embodiment, time spent engaging with the multimedia multi-user collaboration application (e.g., time spent actively speaking, time spent sharing screens, time spent directing gaze toward multimedia multi-user collaboration application, or time spent providing input into multimedia multi-user collaboration application GUI) may be given a positive value between zero and positive one. Time spent not engaging with the multimedia multi-user collaboration application (e.g., user not detected within proximity to information handling system, user muted), or time spent actively engaging with other applications may be given a value between negative one and zero in such an embodiment.).
Regarding claims 3/13: The combination of Swierk and Qiu teaches all the limitations of claims 1/11 above. Swierk doesn’t explicitly teach:
performing, based on values of first-type service features respectively corresponding to the plurality of users, unified encoding on first-type service features having values in a same value range in the plurality of first-type service features, to obtain an encoded first-type service feature;
and performing, based on a type of the second-type service feature, unified encoding on second-type service features belonging to a same category in the plurality of second-type service features, to obtain an encoded second-type service feature;
the aggregating service features in the plurality of time dimensions to obtain a target service feature comprises:
aggregating encoded first-type service features in the plurality of time dimensions, to obtain a target first-type service feature;
and aggregating encoded second-type service features in the plurality of time dimensions, to obtain a target second-type service feature;
and the generating a user feature of the user based on the target account feature and the target service feature comprises generating the user feature of the user based on the target account feature, the target first-type service feature, and the target second-type service feature.
Qiu further teaches:
performing, based on values of first-type service features respectively corresponding to the plurality of users, unified encoding on first-type service features having values in a same value range in the plurality of first-type service features, to obtain an encoded first-type service feature; ([0090] Billing account level features can also include ratios or relative contributions of activity on a billing account by a platform user);
and performing, based on a type of the second-type service feature, unified encoding on second-type service features belonging to a same category in the plurality of second-type service features, to obtain an encoded second-type service feature; ([0090] Billing account level features can also include ratios or relative contributions of activity on a billing account by a platform user);
the aggregating service features in the plurality of time dimensions to obtain a target service feature comprises: aggregating encoded first-type service features in the plurality of time dimensions, to obtain a target first-type service feature; ([0090] billing account features can represent how much of the subscription/purchasing is performed by the platform user versus another user, or a ratio between projects that use subscription services paid for through the billing account that are authored by the platform user versus projects that also use paid-for services but are not authored by the platform user.);
and aggregating encoded second-type service features in the plurality of time dimensions, to obtain a target second-type service feature; ([0090] billing account features can represent how much of the subscription/purchasing is performed by the platform user versus another user, or a ratio between projects that use subscription services paid for through the billing account that are authored by the platform user versus projects that also use paid-for services but are not authored by the platform user.);
and the generating a user feature of the user based on the target account feature and the target service feature comprises generating the user feature of the user based on the target account feature, the target first-type service feature, and the target second-type service feature. ([0092] the platform 110 is configured to maintain a reputation metric that measures a platform user's interaction with different applications and services provided by the platform 110.).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Qiu’s feature(s) listed above. One would’ve been motivated to do so in order to successfully compare platform users according to the metrics (Qiu; [0092]). By incorporating the teachings of Qiu, one would’ve been able to successfully use target features to create user features in order to predict payment likelihood.
Regarding claims 4/14: The combination of Swierk and Qiu teaches all the limitations of claims 1/11 above. Swierk further teaches:
obtaining a plurality of seed users of the application program; ([Column 45, Lines 45-52] meeting metrics may be gathered by the multimedia multi-user collaboration application host server 653, and may describe the number of users, which users are screensharing, which users are using virtual backgrounds, which users are muted, and which participants are hosting, among other descriptions of participation among a plurality of users in a single videoconference session.).
screening the plurality of seed users based on an anomaly indicator of the plurality of seed users; ([Column 45, Lines 45-52] meeting metrics may be gathered by the multimedia multi-user collaboration application host server 653, and may describe the number of users, which users are screensharing, which users are using virtual backgrounds, which users are muted, and which participants are hosting, among other descriptions of participation among a plurality of users in a single videoconference session.).
and determining a seed user obtained through screening as the user of the application program. (([Column 5, Lines 31-43] The trained neural network in embodiments may output optimized media capture instructions for the transmitting information handling system in some embodiments, based on these gathered inputs. For example, optimized media capture instructions for the transmitting information handling system may include instructions to decrease the bit rate at which the microphone captures audio samples, or to decrease the frames per second at which the camera captures video samples, if the transmitting information handling system user is not highly engaged (e.g., speaking often, hosting the session, or screen sharing) in the current user session).
Regarding claims 5/15: The combination of Swierk and Qiu teaches all the limitations of claims 4/14 above. Swierk further teaches:
grouping the plurality of seed users to obtain a plurality of groups; ([Column 45, Lines 45-52] meeting metrics may be gathered by the multimedia multi-user collaboration application host server 653, and may describe the number of users, which users are screensharing, which users are using virtual backgrounds, which users are muted, and which participants are hosting, among other descriptions of participation among a plurality of users in a single videoconference session.);
and determining variance of any one of the plurality of groups to screen the plurality of seed users; ([Column 50, Lines 10-32] the user participation level may represent some combination of the user participation levels described directly above. For example, the user participation levels measured as a percentage of time during the current user session (e.g., time spent muted, time spent directing gaze toward camera, time spent actively speaking or sharing, or time interacting with a GUI for another application) may be represented in decimal form in a scale from negative one to positive one. In such an embodiment, time spent engaging with the multimedia multi-user collaboration application (e.g., time spent actively speaking, time spent sharing screens, time spent directing gaze toward multimedia multi-user collaboration application, or time spent providing input into multimedia multi-user collaboration application GUI) may be given a positive value between zero and positive one. Time spent not engaging with the multimedia multi-user collaboration application (e.g., user not detected within proximity to information handling system, user muted), or time spent actively engaging with other applications may be given a value between negative one and zero in such an embodiment. These various values may be averaged in such an embodiment in order to gauge overall user participation level. In some embodiments, one or more variables included within this determination may be weighted. Examiner notes that one of ordinary skill in the art would reasonably interpret the action to gauge overall participation level of the users as equivalent to determining a variance among the users.);
and the variance determination comprises: obtaining, for one seed user in an ith group of the plurality of groups, duration for which the seed user uses the online meeting software within a first duration range; ([Fig. 8] Step 808 – Determine user participation level for the information handling system of interest.; [Column 50, Lines 12-15] the user participation levels measured as a percentage of time during the current user session (e.g., time spent muted, time spent directing gaze toward camera, time spent actively speaking or sharing, or time interacting with a GUI for another application).);
calculating a variance of use duration of all seed users in the ith group, the ith group being any one of the plurality of groups; ([Column 50, Lines 45-49] The intelligent collaboration contextual session management system may determine in an embodiment whether the user participation levels determined for the information handling system of interest meet preset threshold user participation level requirements at block 810.);
and removing all the seed users in the ith group when the variance is greater than a first variance value; or retaining all the seed users in the ith group when the variance is not greater than a first variance value. ([Column 51, Lines 37-43] If the determined user participation levels for the information handling system of interest meet preset threshold participation level requirements, resolution and quality of audio and video samples may be prioritized, and the method may proceed to block 816 for potential optimization of other participating information handling systems. Examiner notes that one of ordinary skill in the art would reasonably interpret the prioritization of audio and video quality of users that meet a preset threshold as equivalent to retaining all the seed users when the variance is not greater than a first variance value.).
Regarding claims 6/16: The combination of Swierk and Qiu teaches all the limitations of claims 4/14 above. Swierk further teaches:
grouping the plurality of seed users to obtain a plurality of groups; ([Column 45, Lines 45-52] meeting metrics may be gathered by the multimedia multi-user collaboration application host server 653, and may describe the number of users, which users are screensharing, which users are using virtual backgrounds, which users are muted, and which participants are hosting, among other descriptions of participation among a plurality of users in a single videoconference session.);
and determining variance of any one of the plurality of groups to screen the plurality of seed users; ([Column 50, Lines 10-32] the user participation level may represent some combination of the user participation levels described directly above. For example, the user participation levels measured as a percentage of time during the current user session (e.g., time spent muted, time spent directing gaze toward camera, time spent actively speaking or sharing, or time interacting with a GUI for another application) may be represented in decimal form in a scale from negative one to positive one. In such an embodiment, time spent engaging with the multimedia multi-user collaboration application (e.g., time spent actively speaking, time spent sharing screens, time spent directing gaze toward multimedia multi-user collaboration application, or time spent providing input into multimedia multi-user collaboration application GUI) may be given a positive value between zero and positive one. Time spent not engaging with the multimedia multi-user collaboration application (e.g., user not detected within proximity to information handling system, user muted), or time spent actively engaging with other applications may be given a value between negative one and zero in such an embodiment. These various values may be averaged in such an embodiment in order to gauge overall user participation level. In some embodiments, one or more variables included within this determination may be weighted. Examiner notes that one of ordinary skill in the art would reasonably interpret the action to gauge overall participation level of the users as equivalent to determining a variance among the users.);
and the variance determination comprises: obtaining, for one seed user in an ith group of the plurality of groups, a quantity of times that the seed user participates in an online meeting through online meeting software within a second duration range; ([Column 3, Lines 28-35] The intelligent collaboration contextual session management system in embodiments of the present disclosure addresses this issue by training a machine-learning neural network to identify optimized media capture settings and optimized media processing settings for a variety of performance environments encountered by a single information handling system participating in multimedia multi-user collaboration application user sessions over time.; [Column 50, Lines 12-16] the user participation levels measured as a percentage of time during the current user session (e.g., time spent muted, time spent directing gaze toward camera, time spent actively speaking or sharing, or time interacting with a GUI for another application).);
calculating a variance of times of participation of all seed users in the ith group, the ith group being any one of the plurality of groups; ([Column 50, Lines 45-49] The intelligent collaboration contextual session management system may determine in an embodiment whether the user participation levels determined for the information handling system of interest meet preset threshold user participation level requirements at block 810.);
and removing all the seed users in the ith group when the variance is greater than a second variance value; ([Column 50, Lines 49-54] Processing resources consumed by the multimedia multi-user collaboration application at each of the participating information handling systems in an embodiment may be decreased by decreasing the size of the streaming audio and video samples capturing less active users. Examiner notes that one of ordinary skill in the art would reasonably interpret decreasing the size of the streaming audio and video samples capturing less active users as equivalent to removing the less active users when their participation level variance compared to the most active users is greater than a threshold.);
or retaining all the seed users in the ith group when the variance is not greater than a second variance value. ([Column 3, Lines 5-8] Further, commonly only the videos of the most active participants in a large group need this type of high-level processing, or require the capture of high-quality (and thus large data streaming size) video and audio samples.; [Column 48, Lines 43-51] The values gathered at block 804 may identify a number of participants, varying presentation displays, percentage of time users spend on mute, percentage of time users actively engage with the multimedia multi-user collaboration application (e.g., using a messaging service incorporated therewithin), identification of one or more users as a host, or identification of one or more users as sharing screens, for example.).
Claim(s) 7, 8, 17, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Swierk et al. (US 11350058 B1, hereinafter “Swierk”), in view of Qiu et al. (US 20220129318 A1, hereinafter “Qiu”) as applied to claims 1, 4, 11, and 14 above, in further view of Crawford et al. (US 20200320478 A1, hereinafter “Crawford”).
Regarding claims 7/17: The combination of Swierk and Qiu teaches all the limitations of claims 4/14 above. Swierk further teaches:
grouping the plurality of seed users to obtain a plurality of groups; ([Column 45, Lines 45-52] meeting metrics may be gathered by the multimedia multi-user collaboration application host server 653, and may describe the number of users, which users are screensharing, which users are using virtual backgrounds, which users are muted, and which participants are hosting, among other descriptions of participation among a plurality of users in a single videoconference session.);
and determining variance of any one of the plurality of groups to screen the plurality of seed users; ([Column 50, Lines 10-32] the user participation level may represent some combination of the user participation levels described directly above. For example, the user participation levels measured as a percentage of time during the current user session (e.g., time spent muted, time spent directing gaze toward camera, time spent actively speaking or sharing, or time interacting with a GUI for another application) may be represented in decimal form in a scale from negative one to positive one. In such an embodiment, time spent engaging with the multimedia multi-user collaboration application (e.g., time spent actively speaking, time spent sharing screens, time spent directing gaze toward multimedia multi-user collaboration application, or time spent providing input into multimedia multi-user collaboration application GUI) may be given a positive value between zero and positive one. Time spent not engaging with the multimedia multi-user collaboration application (e.g., user not detected within proximity to information handling system, user muted), or time spent actively engaging with other applications may be given a value between negative one and zero in such an embodiment. These various values may be averaged in such an embodiment in order to gauge overall user participation level. In some embodiments, one or more variables included within this determination may be weighted. Examiner notes that one of ordinary skill in the art would reasonably interpret the action to gauge overall participation level of the users as equivalent to determining a variance among the users.);
Swierk doesn’t explicitly teach:
and the variance determination comprises: obtaining, for one seed user in an ith group of the plurality of groups, a duration distribution status in which the seed user participates in an online meeting through online meeting software within a third duration range, the duration distribution status indicating duration for which the seed user participates in a meeting each time and a total quantity of times of participation;
calculating a duration variance of duration for which all seed users in the ith group participate in the meeting each time; calculating a times variance of the total quantity of times of participation of all the seed users in the ith group;
and removing all the seed users in the ith group when the duration variance is greater than a third variance value and the times variance is greater than a fourth variance value; or retaining all the seed users in the ith group when the duration variance is not greater than a third variance value or the times variance is not greater than a fourth variance value.
Crawford teaches:
and the variance determination comprises: obtaining, for one seed user in an ith group of the plurality of groups, a duration distribution status in which the seed user participates in an online meeting through online meeting software within a third duration range, the duration distribution status indicating duration for which the seed user participates in a meeting each time and a total quantity of times of participation; ([0042] In some embodiments, the determined key metrics may include the percentage of participants deemed to be engaged and the percentage of participants deemed to be highly engage. The types, level, and timing of engagement data can be used to categorize participants into segments (e.g., engaged or highly engaged segments) to help group data for reporting and filtering. In some embodiments, segments may overlap. For instance, participants categorized as “Engaged” may include those found “Highly Engaged”. In some embodiments, an “Engaged” participant may be a participant that at a minimum, engaged with the presentation by performing a single click. In some embodiments, a “Highly Engaged Participant” may be a participant that at a minimum, interacted with the presentation by doing more than a single click (e.g., typing a presenter question or note). In some embodiments a metric for “Engaged Percentage” may be generated by determining the number of participants that minimally interacted with the presentation by way of a single click (e.g., a single click may correspond with saving a slide) and divided by the total participants that are logged into the computing devices. Additionally, a metric for a “highly engaged percentage” may be generated by determining the number of participants that more than minimally interacted with the presentation and dividing by the total participants that logged into the devices. For example, this may include determining the number of participants that made a text note, a stylus note, a rating, a submitted presenter question, a response to survey or polling questions or an interaction with a custom button content, or the like.);
calculating a duration variance of duration for which all seed users in the ith group participate in the meeting each time; calculating a times variance of the total quantity of times of participation of all the seed users in the ith group; ([0042] an “Engaged” participant may be a participant that at a minimum, engaged with the presentation by performing a single click. In some embodiments, a “Highly Engaged Participant” may be a participant that at a minimum, interacted with the presentation by doing more than a single click (e.g., typing a presenter question or note). In some embodiments a metric for “Engaged Percentage” may be generated by determining the number of participants that minimally interacted with the presentation by way of a single click (e.g., a single click may correspond with saving a slide) and divided by the total participants that are logged into the computing devices. Additionally, a metric for a “highly engaged percentage” may be generated by determining the number of participants that more than minimally interacted with the presentation and dividing by the total participants that logged into the devices. For example, this may include determining the number of participants that made a text note, a stylus note, a rating, a submitted presenter question, a response to survey or polling questions or an interaction with a custom button content, or the like.; [0044] After categorizing participants, benchmarks can be applied to all the data by calculating average engaged percentage or average highly engaged percentage. The average percentages can be calculated based on the entire data set or subsets (client, series, type of meeting, location, etc.) thus creating benchmarks.);
and removing all the seed users in the ith group when the duration variance is greater than a third variance value and the times variance is greater than a fourth variance value; or retaining all the seed users in the ith group when the duration variance is not greater than a third variance value or the times variance is not greater than a fourth variance value. ([0043] Alternatively, “Engaged” and “Highly Engaged” participants may be determined by the timing of their responses/actions. For example, in some embodiments, a Highly Engaged participant may be defined as one that took actions every 5 minutes or was in the top 10% based on total number of actions. In another example, highly engaged participants may be defined as Or combinations such as answered at least 70% of survey questions and submitted a question. Based on the definitions, a calculation can be applied either in the database or in a BI Tool that can then be used for grouping and filtering data. Examiner notes that one of ordinary skill in the art would reasonably interpret categorizing participants in different groups based on participation as equivalent to both limitations for “removing all seed users…”, and “retaining all seed users…”, as within a group (e.g., engaged), the participants that fall into that category are retained, and the rest are removed.).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Crawford’s feature(s) listed above. One would’ve been motivated to do so in order to successfully correlate and observe relationships between meeting settings and engagement, knowledge attained by participants, and the participant reported experience (Crawford; [0045]). By incorporating the teachings of Crawford, one would’ve been able to successfully use participation variances to categorize meeting participants.
Regarding claims 8/18: The combination of Swierk and Qiu teaches all the limitations of claims 1/11 above. Swierk doesn’t explicitly teach:
generating, based on the plurality of user features, view features of the plurality of users in m views, the view features in different views corresponding to different identifier dimensions of the users;
generating, for a tth view of the m views, a tth predicted payment label matrix based on view features respectively corresponding to the plurality of users in the tth view in combination with a tth projection matrix obtained through pre-training;
calculating a mean of m predicted payment label matrices in the m views; and
determining the mean as the predicted classification result,
the tth projection matrix being configured for aligning the tth view and another view to same projection space, and the another view comprising a view other than the tth view of the m views.
Qiu teaches:
generating, for a tth view of the m views, a tth predicted payment label matrix based on view features respectively corresponding to the plurality of users in the tth view in combination with a tth projection matrix obtained through pre-training; ([0011] The quota resolution system can train and retrain a model that is trained to generate a quota score representing the likelihood that a platform user will abuse additional computing resources. The system can automatically generate new training data from the platform to further train the system in detecting abusive behavior, even as behavioral patterns shift over time. The system can evaluate its own performance according to a variety of different metrics, and retrain itself to more accurately predict abuse by focusing on common characteristics of different subpopulations of a cloud platform user base.; [0092] The platform 110 can generate these reputation metrics separately from the quota resolution system 100, and may use those metrics for a variety of purposes, e.g., for comparing platform users according to the metrics, or for reporting out of the platform 110. For example, one reputation metric for a billing account of the platform user can measure timeliness in making payments to the platform 110. The feature engineering module 215 is configured to use reputation metrics as additional features for generating the engineered set of features, either alone or in combination with other features.);
calculating a mean of m predicted payment label matrices in the m views; and determining the mean as the predicted classification result, ([0022] Generating the score can include generating a predicted revenue likelihood that the computing platform will receive a first request to allocate additional computing resources and a first payment from one of the one or more user accounts within the predetermined future time period, and generating the score as a function comprising the predicted revenue likelihood and the predicted abuse likelihood.; [0096] Because a platform user can be represented by different user accounts, the feature engineering module 215 can aggregate any of the above-described features to represent an aggregated feature across all user accounts representing the platform user. The feature engineering module 215 can receive as input a mapping of identifiers representing each user account of the platform user and how each account relates to one another. For example, if the platform user is represented by several billing accounts, then the feature engineering module 215 can aggregate payment history across each account, creating an aggregated payment history feature. The value for the feature in this case can be an average payment history across billing accounts for the platform user, as an example. FIG. 3, below, shows an example of the mapping.; [0105] In some implementations, the feature engineering module 215 uses data across a plurality of time-steps corresponding to the platform user to determine statistical trends. For example, the feature engineering module 215 can generate a value for a feature representing a trend of the platform user's resource consumption over a period of time. This and other features can then be processed by the scoring-serving model to predict future behavior, i.e., the likelihood of abuse by the platform user given additional resources. In some implementations, the feature engineering module 215 uses data across a plurality of different platform users to generate trends or baselines from which to compare with feature data.).
the tth projection matrix being configured for aligning the tth view and another view to same projection space, and the another view comprising a view other than the tth view of the m views. ([0092] Reputation is a general type of feature that can be available across user accounts representing the platform user. In some implementations, the platform 110 is configured to maintain a reputation metric that measures a platform user's interaction with different applications and services provided by the platform 110. The platform 110 can generate these reputation metrics separately from the quota resolution system 100, and may use those metrics for a variety of purposes, e.g., for comparing platform users according to the metrics, or for reporting out of the platform 110. For example, one reputation metric for a billing account of the platform user can measure timeliness in making payments to the platform 110. The feature engineering module 215 is configured to use reputation metrics as additional features for generating the engineered set of features, either alone or in combination with other features. Examiner notes that one of ordinary skill in the art would reasonably interpret the different metrics reported, including metrics related to payments, as equivalent to aligning tth view to another view.).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Qiu’s feature(s) listed above. One would’ve been motivated to do so in order to successfully maintain a reputation metric that measures a platform user's interaction with different applications and services provided by the platform 110 (Qiu; [0092]). By incorporating the teachings of Qiu, one would’ve been able to successfully calculate a mean used to predict behavior.
Crawford teaches:
generating, based on the plurality of user features, view features of the plurality of users in m views, the view features in different views corresponding to different identifier dimensions of the users; ([0021] In some embodiments, the systems and methods built in accordance with the present disclosure may analyze audience engagement with a presentation displayed during a live meeting and may provide feedback to a presenter during and/or after the presentation. In some embodiments, the provided feedback may include determining one or more metrics indicative of participant engagement levels. Participant engagement levels may then be used to adjust live meetings in real-time and/or after the live meeting has ended.);
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Crawford’s feature(s) listed above. One would’ve been motivated to do so, so that engagement metrics 529 may also be displayed (Crawford; [0065]). By incorporating the teachings of Crawford, one would’ve been able to successfully view identifier dimension metrics from the plurality of users.
Claim(s) 9, 10, and 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over Swierk et al. (US 11350058 B1, hereinafter “Swierk”), in view of Qiu et al. (US 20220129318 A1, hereinafter “Qiu”) as applied to claims 1/11 above, in further view of Crawford et al. (US 20200320478 A1, hereinafter “Crawford”), in further view of Pang et al. (US 12027151 B2, hereinafter “Pang”).
Regarding claims 9/19: The combination of Swierk and Qiu teaches all the limitations of claims 1/11 above. Swierk doesn’t teach:
inputting, for one of the plurality of user features, the user feature into an encoder, to obtain a low-dimensional user feature;
inputting the low-dimensional user feature into a decoder, to obtain a high-dimensional user feature;
adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition;
determining the high-dimensional user feature as the user feature;
re-performing the inputting the user feature into an encoder, to obtain a low-dimensional user feature;
and calculating an error between the user feature and the high-dimensional user feature; or outputting the low-dimensional user feature when the error reaches a convergence condition;
and clustering the plurality of users based on a plurality of low-dimensional user features corresponding to the plurality of user features, to obtain the predicted classification result.
Pang teaches:
inputting, for one of the plurality of user features, the user feature into an encoder, to obtain a low-dimensional user feature; ([Abstract] The content encoder is configured to receive input speech as input and generate a latent representation of linguistic content for the input speech output. The content encoder is trained to disentangle speaking style information from the latent representation of linguistic content. The style encoder is configured to receive the input speech as input and generate a latent representation of speaking style for the input speech as output. The style encoder is trained to disentangle linguistic content information from the latent representation of speaking style. The decoder is configured to generate output speech based on the latent representation of linguistic content for the input speech and the latent representation of speaking style for the same or different input speech. Examiner notes that one of ordinary skill in the art would reasonably consider the speaking style as low-dimensional target user feature.);
inputting the low-dimensional user feature into a decoder, to obtain a high-dimensional user feature; ([Column 5, Lines 11-16] The autoencoder model includes a content encoder, a style encoder 130, and a decoder. The decoder is configured to receive both content and style latent representations as input, and generate speech features as output. That is, the decoder is configured to reconstruct the input speech as the output speech features.; [Column 8, Line52-Column 9, Line 11] Experiments show that increasing a VQ-VAE codebook size of the content encoder 110 optimizes the model 100 for preserving linguistic content from input speech 102. Additionally, applying the mutual information loss to minimize mutual information captured by the content and style encoders 110, 130 further improves linguistic content preservation. In a first non-shuffle scenario, optimized for measuring how well the model 100 compresses speech, the content and style encoders 110, 130 each receive the same input speech 102, X.sub.i, and the decoder 150 predicts speech features 152, {circumflex over (X)}, that correspond to a reconstruction of the input speech.; [Column 5, Lines 44-51] During inference, the computing system 20 (e.g., the data processing hardware 22) or a user computing device (not shown) executes the trained autoencoder model 100 that includes a content encoder 110, a style encoder 130, and a decoder 150 to generate new speech features 152 as synthesized speech that conveys the linguistic content extracted from a first speech sample 50, 50a and having a speaking style extracted from a second speech sample 50, 50b. Examiner notes that one of ordinary skill in the art would reasonably consider the speaking style received by the decoder as a low-dimensional feature, and speech features as high-dimensional features.);
adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition; ([Column 2, Lines 14-19] the style encoder may be trained using a style regularization loss based on a mean and variance of style latent variables predicted by the style encoder, wherein the style encoder uses the style regularization loss to minimize a Kullback-Leibler (KL) divergence between a Gaussian posterior with a unit Gaussian prior.; [Column 2, Lines 20-27] the decoder is configured to: receive, as input, the latent representation of linguistic content for the input speech and the latent representation of speaking style for the same input speech; and generate, as output, the output speech comprising a reconstruction of the input speech. The model may be trained using a reconstruction loss between the input speech and the reconstruction of the input speech output from the decoder. Examiner notes that one of ordinary skill in the art would reasonably interpret the use of regularization loss and reconstruction loss in training as equivalent to adjusting the encoder and decoder until it reaches convergence.);
determining the high-dimensional user feature as the user feature; ([Column 4, Lines 19-29] Speech waveforms are a complex, high-dimensional form of data influenced by a number of underlying factors, which can be broadly categorized into linguistic contents (e.g., phonemes) and speaking styles. Learning disentangled latent representations of linguistic content and speaking style from speech has a wide set of applications in generative tasks, including speech synthesis, data augmentation, voice transfer, and speech compression. Disentangling latent representations from speech can also be helpful for downstream tasks such as automated speech recognition and speaker classification.);
re-performing the inputting the user feature into an encoder, to obtain a low-dimensional user feature; ([Column 10, Lines 38-54] FIG. 4B shows another encoder-only application that includes using the trained content encoder 110 as a speech recognition model 400b for generating speech recognition results for input speech 402. The content encoder 110 may be leveraged to extract latent representations of linguistic content 120 to provide local information for use in bootstrapping automated speech recognition (ASR) training by using the latent representations of linguistic content 120 and leveraging unsupervised data. In the example shown, a neural network 470 is overlain on top of the content encoder 110 to provide the speech recognition model 400b. In some implementations, the content encoder 110 and the neural network 470 are retrained using labeled data of input speech examples 402 and corresponding transcriptions. In this arrangement, the content encoder 110 functions as a feature extractor for encoding linguistic content from speech for improving speech recognition accuracy.);
and calculating an error between the user feature and the high-dimensional user feature; or outputting the low-dimensional user feature when the error reaches a convergence condition; ([Column 9, Lines 2-8] To evaluate how well the predicted speech features output by the decoder compared to the original linguistic content from the input speech, an automated speech recognizer transcribed the predicted speech features {circumflex over (X)} and a word error rate is calculated for the transcription with ground-truth text for the original input speech X.sub.i fed to the content encoder 110.);
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Pang’s feature(s) listed above. One would’ve been motivated to do so in order to successfully summarize the output from the convolution layer 132 with a global average pooling operation across the time-axis (Pang; [Column 7, Lines 7-8]). By incorporating the teachings of Pang, one would’ve been able to successfully train encoders and decoders in the model to obtain low-dimensional and high-dimensional user features.
Pang doesn’t teach:
and clustering the plurality of users based on a plurality of low-dimensional user features corresponding to the plurality of user features, to obtain the predicted classification result.
Crawford teaches:
and clustering the plurality of users based on a plurality of low-dimensional user features corresponding to the plurality of user features, to obtain the predicted classification result. ([0042] In some embodiments, the determined key metrics may include the percentage of participants deemed to be engaged and the percentage of participants deemed to be highly engage. The types, level, and timing of engagement data can be used to categorize participants into segments (e.g., engaged or highly engaged segments) to help group data for reporting and filtering. In some embodiments, segments may overlap. For instance, participants categorized as “Engaged” may include those found “Highly Engaged”. In some embodiments, an “Engaged” participant may be a participant that at a minimum, engaged with the presentation by performing a single click. In some embodiments, a “Highly Engaged Participant” may be a participant that at a minimum, interacted with the presentation by doing more than a single click (e.g., typing a presenter question or note). In some embodiments a metric for “Engaged Percentage” may be generated by determining the number of participants that minimally interacted with the presentation by way of a single click (e.g., a single click may correspond with saving a slide) and divided by the total participants that are logged into the computing devices. Additionally, a metric for a “highly engaged percentage” may be generated by determining the number of participants that more than minimally interacted with the presentation and dividing by the total participants that logged into the devices. For example, this may include determining the number of participants that made a text note, a stylus note, a rating, a submitted presenter question, a response to survey or polling questions or an interaction with a custom button content, or the like.).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Crawford’s feature(s) listed above. One would’ve been motivated to do so in order to successfully group and filter data (Crawford; [0043]). By incorporating the teachings of Crawford, one would’ve been able to successfully cluster meeting participant based on their features.
Regarding claims 10: The combination of Swierk and Qiu teaches all the limitations of claim 1 above. Swierk doesn’t teach:
generating, based on the plurality of user features, view features of the plurality of users in m views;
obtaining, through training for a tth view of the m views, a tth projection matrix based on view features respectively corresponding to the plurality of users in the tth view and payment labels of some of the plurality of users;
generating a target projection matrix based on m projection matrices corresponding to the m views, the target projection matrix being configured for indicating a plurality of target user features respectively corresponding to the plurality of users;
inputting, for one of the plurality of target user features, the target user feature into an encoder, to obtain a low-dimensional target user feature;
inputting the low-dimensional target user feature into a decoder, to obtain a high-dimensional target user feature;
adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition;
determining the high-dimensional user feature as the target user feature;
re-performing the inputting the target user feature into an encoder, to obtain a low-dimensional target user feature;
and calculating an error between the target user feature and the high-dimensional target user feature; or outputting the low-dimensional target user feature when the error reaches a convergence condition;
and clustering the plurality of users based on a plurality of low-dimensional target user features corresponding to the plurality of target user features, to obtain the predicted classification result,
the tth projection matrix being configured for aligning the tth view and another view to same projection space, and the another view comprising a view other than the tth view of the m views.
Pang teaches:
inputting, for one of the plurality of target user features, the target user feature into an encoder, to obtain a low-dimensional target user feature; ([Abstract] The content encoder is configured to receive input speech as input and generate a latent representation of linguistic content for the input speech output. The content encoder is trained to disentangle speaking style information from the latent representation of linguistic content. The style encoder is configured to receive the input speech as input and generate a latent representation of speaking style for the input speech as output. The style encoder is trained to disentangle linguistic content information from the latent representation of speaking style. The decoder is configured to generate output speech based on the latent representation of linguistic content for the input speech and the latent representation of speaking style for the same or different input speech. Examiner notes that one of ordinary skill in the art would reasonably consider the speaking style as low-dimensional target user feature.);
inputting the low-dimensional target user feature into a decoder, to obtain a high-dimensional target user feature; ([Column 5, Lines 11-16] The autoencoder model includes a content encoder, a style encoder 130, and a decoder. The decoder is configured to receive both content and style latent representations as input, and generate speech features as output. That is, the decoder is configured to reconstruct the input speech as the output speech features.; [Column 8, Line52-Column 9, Line 11] Experiments show that increasing a VQ-VAE codebook size of the content encoder 110 optimizes the model 100 for preserving linguistic content from input speech 102. Additionally, applying the mutual information loss to minimize mutual information captured by the content and style encoders 110, 130 further improves linguistic content preservation. In a first non-shuffle scenario, optimized for measuring how well the model 100 compresses speech, the content and style encoders 110, 130 each receive the same input speech 102, X.sub.i, and the decoder 150 predicts speech features 152, {circumflex over (X)}, that correspond to a reconstruction of the input speech.; [Column 5, Lines 44-51] During inference, the computing system 20 (e.g., the data processing hardware 22) or a user computing device (not shown) executes the trained autoencoder model 100 that includes a content encoder 110, a style encoder 130, and a decoder 150 to generate new speech features 152 as synthesized speech that conveys the linguistic content extracted from a first speech sample 50, 50a and having a speaking style extracted from a second speech sample 50, 50b. Examiner notes that one of ordinary skill in the art would reasonably consider the speaking style received by the decoder as a low-dimensional feature, and speech features as high-dimensional features.);
adjusting the encoder and the decoder based on the error when the error does not reach a convergence condition; ([Column 2, Lines 14-19] the style encoder may be trained using a style regularization loss based on a mean and variance of style latent variables predicted by the style encoder, wherein the style encoder uses the style regularization loss to minimize a Kullback-Leibler (KL) divergence between a Gaussian posterior with a unit Gaussian prior.; [Column 2, Lines 20-27] the decoder is configured to: receive, as input, the latent representation of linguistic content for the input speech and the latent representation of speaking style for the same input speech; and generate, as output, the output speech comprising a reconstruction of the input speech. The model may be trained using a reconstruction loss between the input speech and the reconstruction of the input speech output from the decoder. Examiner notes that one of ordinary skill in the art would reasonably interpret the use of regularization loss and reconstruction loss in training as equivalent to adjusting the encoder and decoder until it reaches convergence.);
determining the high-dimensional user feature as the target user feature; ([Column 4, Lines 19-29] Speech waveforms are a complex, high-dimensional form of data influenced by a number of underlying factors, which can be broadly categorized into linguistic contents (e.g., phonemes) and speaking styles. Learning disentangled latent representations of linguistic content and speaking style from speech has a wide set of applications in generative tasks, including speech synthesis, data augmentation, voice transfer, and speech compression. Disentangling latent representations from speech can also be helpful for downstream tasks such as automated speech recognition and speaker classification.);
re-performing the inputting the target user feature into an encoder, to obtain a low-dimensional target user feature; ([Column 10, Lines 38-54] FIG. 4B shows another encoder-only application that includes using the trained content encoder 110 as a speech recognition model 400b for generating speech recognition results for input speech 402. The content encoder 110 may be leveraged to extract latent representations of linguistic content 120 to provide local information for use in bootstrapping automated speech recognition (ASR) training by using the latent representations of linguistic content 120 and leveraging unsupervised data. In the example shown, a neural network 470 is overlain on top of the content encoder 110 to provide the speech recognition model 400b. In some implementations, the content encoder 110 and the neural network 470 are retrained using labeled data of input speech examples 402 and corresponding transcriptions. In this arrangement, the content encoder 110 functions as a feature extractor for encoding linguistic content from speech for improving speech recognition accuracy.);
and calculating an error between the target user feature and the high-dimensional target user feature; or outputting the low-dimensional target user feature when the error reaches a convergence condition; ([Column 9, Lines 2-8] To evaluate how well the predicted speech features output by the decoder compared to the original linguistic content from the input speech, an automated speech recognizer transcribed the predicted speech features {circumflex over (X)} and a word error rate is calculated for the transcription with ground-truth text for the original input speech X.sub.i fed to the content encoder 110.).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Pang’s feature(s) listed above. One would’ve been motivated to do so in order to successfully summarize the output from the convolution layer 132 with a global average pooling operation across the time-axis (Pang; [Column 7, Lines 7-8]). By incorporating the teachings of Pang, one would’ve been able to successfully train encoders and decoders in the model to obtain low-dimensional and high-dimensional user features.
Pang doesn’t teach:
generating, based on the plurality of user features, view features of the plurality of users in m views;
obtaining, through training for a tth view of the m views, a tth projection matrix based on view features respectively corresponding to the plurality of users in the tth view and payment labels of some of the plurality of users;
generating a target projection matrix based on m projection matrices corresponding to the m views, the target projection matrix being configured for indicating a plurality of target user features respectively corresponding to the plurality of users;
and clustering the plurality of users based on a plurality of low-dimensional target user features corresponding to the plurality of target user features, to obtain the predicted classification result,
the tth projection matrix being configured for aligning the tth view and another view to same projection space, and the another view comprising a view other than the tth view of the m views.
Qiu teaches:
obtaining, through training for a tth view of the m views, a tth projection matrix based on view features respectively corresponding to the plurality of users in the tth view and payment labels of some of the plurality of users; ([0011] The quota resolution system can train and retrain a model that is trained to generate a quota score representing the likelihood that a platform user will abuse additional computing resources. The system can automatically generate new training data from the platform to further train the system in detecting abusive behavior, even as behavioral patterns shift over time. The system can evaluate its own performance according to a variety of different metrics, and retrain itself to more accurately predict abuse by focusing on common characteristics of different subpopulations of a cloud platform user base.; [0016] Generating the score can include generating the score using one or more machine learning models trained to generate the score from the plurality of features. The operations can further include determining, by the system, that the score does not meet a plurality of evaluation criteria; and in response to determining that the score does not meet the plurality of evaluation criteria, retraining the one or more machine learning models trained to generate the score from the plurality of features.; [0092] The platform 110 can generate these reputation metrics separately from the quota resolution system 100, and may use those metrics for a variety of purposes, e.g., for comparing platform users according to the metrics, or for reporting out of the platform 110. For example, one reputation metric for a billing account of the platform user can measure timeliness in making payments to the platform 110. The feature engineering module 215 is configured to use reputation metrics as additional features for generating the engineered set of features, either alone or in combination with other features.);
generating a target projection matrix based on m projection matrices corresponding to the m views, the target projection matrix being configured for indicating a plurality of target user features respectively corresponding to the plurality of users; ([0106] The feature engineering module 215 can include one or more machine learning models trained to receive data from the data orchestration module 205 and to generate values for one or more features from the received data. In this way, the feature engineering module 215 can further extract potential candidates for the engineered set of features, leveraging the data-rich nature of the platform 110.);
the tth projection matrix being configured for aligning the tth view and another view to same projection space, and the another view comprising a view other than the tth view of the m views. ([0092] Reputation is a general type of feature that can be available across user accounts representing the platform user. In some implementations, the platform 110 is configured to maintain a reputation metric that measures a platform user's interaction with different applications and services provided by the platform 110. The platform 110 can generate these reputation metrics separately from the quota resolution system 100, and may use those metrics for a variety of purposes, e.g., for comparing platform users according to the metrics, or for reporting out of the platform 110. For example, one reputation metric for a billing account of the platform user can measure timeliness in making payments to the platform 110. The feature engineering module 215 is configured to use reputation metrics as additional features for generating the engineered set of features, either alone or in combination with other features. Examiner notes that one of ordinary skill in the art would reasonably interpret the different metrics reported, including metrics related to payments, as equivalent to aligning tth view to another view.).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Qiu’s feature(s) listed above. One would’ve been motivated to do so in order to successfully maintain a reputation metric that measures a platform user's interaction with different applications and services provided by the platform 110 (Qiu; [0092]). By incorporating the teachings of Qiu, one would’ve been able to successfully calculate a mean used to predict behavior.
Qiu doesn’t teach:
generating, based on the plurality of user features, view features of the plurality of users in m views;
and clustering the plurality of users based on a plurality of low-dimensional target user features corresponding to the plurality of target user features, to obtain the predicted classification result,
Crawford teaches:
generating, based on the plurality of user features, view features of the plurality of users in m views; ([0021] In some embodiments, the systems and methods built in accordance with the present disclosure may analyze audience engagement with a presentation displayed during a live meeting and may provide feedback to a presenter during and/or after the presentation. In some embodiments, the provided feedback may include determining one or more metrics indicative of participant engagement levels. Participant engagement levels may then be used to adjust live meetings in real-time and/or after the live meeting has ended.);
and clustering the plurality of users based on a plurality of low-dimensional target user features corresponding to the plurality of target user features, to obtain the predicted classification result, ([0042] In some embodiments, the determined key metrics may include the percentage of participants deemed to be engaged and the percentage of participants deemed to be highly engage. The types, level, and timing of engagement data can be used to categorize participants into segments (e.g., engaged or highly engaged segments) to help group data for reporting and filtering. In some embodiments, segments may overlap. For instance, participants categorized as “Engaged” may include those found “Highly Engaged”. In some embodiments, an “Engaged” participant may be a participant that at a minimum, engaged with the presentation by performing a single click. In some embodiments, a “Highly Engaged Participant” may be a participant that at a minimum, interacted with the presentation by doing more than a single click (e.g., typing a presenter question or note). In some embodiments a metric for “Engaged Percentage” may be generated by determining the number of participants that minimally interacted with the presentation by way of a single click (e.g., a single click may correspond with saving a slide) and divided by the total participants that are logged into the computing devices. Additionally, a metric for a “highly engaged percentage” may be generated by determining the number of participants that more than minimally interacted with the presentation and dividing by the total participants that logged into the devices. For example, this may include determining the number of participants that made a text note, a stylus note, a rating, a submitted presenter question, a response to survey or polling questions or an interaction with a custom button content, or the like.).
It would have been obvious to one of ordinary skill in the art, at the time of applicant’s invention, to combine modified Swierk with Crawford’s feature(s) listed above. One would’ve been motivated to do so, so that engagement metrics 529 may also be displayed (Crawford; [0065]), and group and filter data (Crawford; [0043]). By incorporating the teachings of Crawford, one would’ve been able to successfully view identifier dimension metrics from the plurality of users, and cluster meeting participant based on their features.
Conclusion
The following prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Peters et al. (WO 2022046168 A1), which discloses a system for processing a stream of user state data indicating attributes of a user of the endpoint device at different times during the network-based communication session.
B. B. S. Vaishnavi, B. Harsha, N. N. Chandana, V. Bhargavi and S. Manne, "Online Video Conferencing with Report Generation," 2021 Fifth International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC), Palladam, India, 2021, pp. 850-856, which discloses using conferencing apps to connect students and teachers, and to conduct online classes.
Chen et al. Analysis of User Satisfaction with Online Education Platforms in China during the COVID-19 Pandemic. Healthcare (Basel). 2020 Jul 7, which discloses using a questionnaire survey and web crawler to collect experience data of online and offline users.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GABRIEL J TORRES CHANZA whose telephone number is (571)272-3701. The examiner can normally be reached Monday thru Friday 8am - 5pm ET.
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, Brian Epstein can be reached on (571)270-5389. 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.
/G.J.T./Examiner, Art Unit 3625
/SARA GRACE BROWN/Primary Examiner, Art Unit 3625
Prosecution Insights