DETAILED ACTION
Notice of Pre-AIA or AIA Status
Claims 1-20 are presented for examination.
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 103
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
Determining the scope and contents of the prior art.
Ascertaining the differences between the prior art and the claims at issue.
Resolving the level of ordinary skill in the pertinent art.
Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 4, 6, 9, 11-12, 15 and 17-19 are being rejected under 35 U.S.C. 103 as being unpatentable over Ie et al (RECSIM: A Configurable Simulation Platform for Recommender Systems [See IDS, 11/04/2022]), hereinafter referred to as Ie.
Regarding Claim 1, Ie discloses
“A system for providing a simulator augmented content selection”, (See Introduction, pg. 2, last paragraph, wherein it is discussed that RecSim is a simulator platform for with a controlled environment for developing recommender models and algorithms.)
“a content selection object generator configured to generate a content selection object corresponding to a candidate content selection machine learning model trained to predict one or more selectable content media items for at least one simulated user”. (See section 3.1 paragraphs 1-3, and section 3.2, wherein Ie discloses a recommender agent configured to generate slates of documents for simulated users. The agent, embodied as a machine learning model such as FullSlateQAgent (a DQN- based model) or TabularQAgent – is trained to predict and select one or more documents (content media items) for presentation to at least one simulated user. The documents explicitly include media content such as music tracks, videos, and news articles. In the interest of compact prosecution, also note that the applicant’s specification confirms that the claimed principles apply broadly beyond traditional media content, therefore the documents disclosed can be distinguished on the basis of being “media items”. (Also see Ie, Page 12, 4.1, last paragraph, wherein the current environment design uses a simulator that sequentially simulates each user.)
“a simulated user model selector configured to provide a simulated user model corresponding to a simulated user model trained to predict a next action of the at least one simulated user in response to a simulated playback input received from the at least one simulated user”. (See Ie Page 9, Section 3.1, and Page 12, Section 4.1, wherein the environment constitutes to a simulated user model selector, which at the beginning of each episode, establishes the environment as a selector that provides the appropriate user model. Multiple user model configuration including multinomial logit and exponentiated cascade choice models also confirms the selector functionality. Simulated user model is also disclosed as the user choice model which predicts the user’s next action in response to presented content. Moreover, the recommender slate constitutes the simulated playback input. The recommender slate are the documents presented to the simulated users at each turn. This being the content stimulus that the user choice model processes to generate a predicted response. The feedback loop, where prior user actions update both the user state and the agent policy.)
“a session initializer configured to initialize the content selection object according to a plurality of session initialization parameter values, thereby generating an initialized content selection object, where the plurality of session initialization parameter values corresponds to a simulated media content playback session”. (See at least Section 4.1, Page 12, wherein Ie establishes the initial user state parameters for the simulated session. The episode initialization represents a session initializer. “At each iteration the simulator aggregates and logs relevant metrics” which speak for the plurality of session initialization parameter values. Moreover, the recommender slate constitutes the simulated playback input. The recommender slate are the documents presented to the simulated users at each turn. This being the content stimulus that the user choice model processes to generate a predicted response. The feedback loop, where prior user actions update both the user state and the agent policy.)
“an augmented content selection simulator configured to apply the simulated user model to content items identified by the initialized content selection object to generate a simulated session including a sequence of predicted simulated user next actions and one or more predicted sets of selectable content items, each set of the one or more selectable content items correlated to each next action in the sequence of predicted simulated user next actions.” (See at least Section 3.1, 4.1 and Figure 2) While Ie doesn’t explicitly disclose an “augmented content selection simulator”, the simulator applied to the user choice model constitutes the simulated user model, and documents identified by the agent constitutes the initialized content selection object, across multiple turns within an episode. The disclosed agent selecting a slate at each turn constitutes the predicted set of selectable content items, and the user choice model generating a predicted response constitutes the predicted next action, producing a turn-by-turn correlated sequence of content sets and user actions across the session. Furthermore, Ie’s RecSim is characterized as a dynamic Bayesian network defining a probability distribution over trajectories of correlated slates and user choices. Therefore, it would have been obvious to one of ordinary skill in the art at the time of the invention was filed to apply the disclosed episode structure, control flow, and trajectory formulation to an augmented content selection system in order to generate correlated sequences within the content selection domain. Ie recognizes that simulation augmented content selection is necessary to safely develop and evaluate recommendation algorithms without exposing real users to experimental systems. (See section 1).
Regarding Claim 4, Ie discloses
“The system according to claim 1, further comprising: a candidate content selection simulator configured to register a plurality of simulated user models, wherein the simulated user model is obtained by a selection of one of the plurality of simulated user models, wherein each simulated user model is trained to uniquely predict a next action of the at least one simulated user in response to a simulated playback input received from the at least one simulated user.” (See at least Section 4.1 and 4.3, wherein the registration of a plurality of simulated user models constitutes multiple configurable user model implementations. Also note the hierarchical agent layers, wherein different user model configurations produce distinct predicted responses to the same presented content, constituting simulated user models being trained to uniquely predict a next action. While Ie does not disclose “Candidate content selection simulator” by name, it would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to implement the plurality of selectable, distinctly configured user models as a registration and selection mechanism, as this represents multi-model framework into a formal registration system.) Ie recognizes that registering and comparing multiple candidate models within a unified simulation platform is a necessary and expected component of the content selection development workflow. (See Section 1.4)
Regarding Claim 6, Ie discloses
“The system of claim 1, the simulated user model selector further configured to: generate the simulated user model by applying any one of (i) a plurality of predefined attributes of the one or more simulated users, (ii) a plurality of predefined attributes of media content items, or (iii) a combination of (i) and (ii), to the selected simulated user model.” (See Section 3.1 and Page 4, paragraph, wherein Ie discloses that the user model samples users from a prior distribution over configurable user features, including latent features such as personality, satisfaction and interests along with observable features such as demographics, and behavioral features such as session length and visit frequency. The predefined user feature distributions constitute the plurality of predefined attributes of the one or more simulated users. While Ie does not disclose exactly the “predefined attributes” verbiage, It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to generate a simulated user model by applying these predefined user and document features distributions to configure realistic simulated user behavior). Ie recognizes that a formal input/output interface through which operators drive model selection and configuration is an expected component of the simulation framework. (See Section 4.1).
Regarding Claim 9, Ie discloses
“A method for providing a simulator augmented content selection, comprising the steps of: receiving a content selection object corresponding to a candidate content selection machine learning model trained to predict one or more selectable content media items for at least one simulated user;”. (See section 3.1 and 4.1, wherein it is disclosed that the recommender slates of documents to a user which represents a content selection object generator generating a content selection object. The recommender agent embodied as a machine learning model such as FullSlateQAgent or TabularQAgent is configured to select or recommend slates of documents for at least one simulated user. The receipt of a content selection object corresponding to a candidate ML model trained to predict selectable content items.)
“receiving a simulated user model corresponding to a simulated user model trained to predict a next action of the at least one simulated user in response to a simulated playback input received from the at least one simulated user;” (See Section 4.1, wherein Ie discloses that the receiver receives and applies a user model for each simulated episode. The simulator asks the environment to sample a user model at the beginning of each episode, suggesting the receipt of a simulated user model. The user choice model within that user model is trained to predict the user’s next action in response to the recommended slate. Also note the feedback loop, wherein prior user actions update both user state and agent policy, suggesting the characterization of the playback inputs as being received from the simulated user.)
“initializing the content selection object according to a plurality of session initialization parameter values, thereby generating an initialized content selection object, where the plurality of session initialization parameter values correspond to a simulated media content playback session”. (See at least Section 4.1, wherein Ie establishing initial user state parameters for the simulated session. The episode initialization represents a session initializer. “At each iteration the simulator aggregates and logs relevant metrics” which speak for the plurality of session initialization parameter values. The simulator initializes the user model, document model, and agent with configurable parameters including user features, document features and session dynamics. These initialization parameters collectively suggest the plurality of session initialization parameter values.)
“and applying the simulated user model to content items identified by the initialized content selection object to generate a simulated session including a sequence of predicted simulated user next actions and one or more predicted sets of selectable content items, each set of the one or more selectable content items correlated to each next action in the sequence of predicted simulated user next actions.” (See at least Section 4.1, and Fig 2, wherein Ie discloses the multi-turn simulation loop, wherein the simulator applies the user choice model to documents selected by the agent across multiple turns of an episode.) While the certain verbiage might differ in Ie’s teaching, it is directly related to the same field of recommender system simulation and address the same problem of safely testing content selection algorithms without exposing real users to experimental systems. Ie explicitly discloses that the goal is to “reduce live experiment cycle time via rapid development and model refinement in simulation, and minimizes the potential for negative impact on users in real-world systems.” (See Introduction). It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to implement the simulation architecture to produce a method for augmented content selection using simulated users. Ie recognizes that a simulation-based method for content selection is necessary to protect real users from experimental algorithms while enabling rapid development and evaluation of recommendation models. (See Section 1).
Regarding Claim 11,
“The method according to claim 9, further comprising: registering a plurality of simulated user models; receiving a selection of one of the plurality of simulated user models, to obtain the simulated user model, wherein each simulated user model is trained to uniquely predict a next action of the at least one simulated user in response to a simulated playback input received from the at least one simulated user.” The claim claims the method counterparts corresponding to the system claim 4 above, and is rejected, at least for the same reasons as claim 4.
Regarding Claim 12,
“The method of claim 9, further comprising: generating the simulated user model by applying any one of (i) a plurality of predefined attributes of the one or more simulated users, (ii) a plurality of predefined attributes of media content items, or (iii) a combination of (i) and (ii), to the selected simulated user model.” The claim claims the method counterparts corresponding to the system claim 6 above, and is rejected, at least for the same reasons as claim 6.
Regarding Claim 15, Ie discloses
“A non-transitory computer-readable medium having stored thereon one or more sequences of instructions for causing one or more processors to perform: and receiving a content selection object corresponding to a candidate content selection machine learning model trained to predict one or more selectable content media items for at least one simulated user”. (See section 3.1 and 4.1, wherein Ie discloses relying on Tensorflow, and logging episodes as Tensorflow Sequence Examples confirming that the instructions are stored and executed on computer-readable medium. Furthermore, the recommender slates of documents to a user which represents a content selection object generator generating a content selection object. The recommender agent embodied as a machine learning model such as FullSlateQAgent or TabularQAgent is configured to select or recommend slates of documents for at least one simulated user. The receipt of a content selection object corresponding to a candidate ML model trained to predict selectable content items.)
“receiving a simulated user model corresponding to a simulated user model trained to predict a next action of the at least one simulated user in response to a simulated playback input received from the at least one simulated user”. (See Section 4.1, wherein Ie discloses that the receiver receives and applies a user model for each simulated episode. The simulator asks the environment to sample a user model at the beginning of each episode, suggesting the receipt of a simulated user model. The user choice model within that user model is trained to predict the user’s next action in response to the recommended slate. Also note the feedback loop, wherein prior user actions update both user state and agent policy, suggesting the characterization of the playback inputs as being received from the simulated user.)
“initializing the content selection object according to a plurality of session initialization parameter values, thereby generating an initialized content selection object, where the plurality of session initialization parameter values correspond to a simulated media content playback session”. (See at least Section 4.1, wherein Ie establishing initial user state parameters for the simulated session. The episode initialization represents a session initializer. “At each iteration the simulator aggregates and logs relevant metrics” which speak for the plurality of session initialization parameter values. The simulator initializes the user model, document model, and agent with configurable parameters including user features, document features and session dynamics. These initialization parameters collectively suggest the plurality of session initialization parameter values.)
“and applying the simulated user model to content items identified by the initialized content selection object to generate a simulated session including a sequence of predicted simulated user next actions and one or more predicted sets of selectable content items, each set of the one or more selectable content items correlated to each next action in the sequence of predicted simulated user next actions.” (See at least Section 4.1, and Fig 2, wherein Ie discloses the multi-turn simulation loop, wherein the simulator applies the user choice model to documents selected by the agent across multiple turns of an episode.) While the certain verbiage might differ in Ie’s teaching, it is directly related to the same field of recommender system simulation and address the same problem of safely testing content selection algorithms without exposing real users to experimental systems. Ie explicitly discloses that the goal is to “reduce live experiment cycle time via rapid development and model refinement in simulation, and minimizes the potential for negative impact on users in real-world systems.” (See Introduction). It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to implement the simulation architecture to produce a method for augmented content selection using simulated users. As Ie’s RecSim is explicitly designed and distributed as stored processor-executable instructions on computer-readable media to achieve its stated goals of reproducibility and sharing within the research community. (See Section 1.4).
Regarding Claim 17, Ie discloses
“The non-transitory computer-readable medium of claim 15, further having stored thereon a sequence of instructions for causing the one or more processors to perform: registering a plurality of simulated user models; and receiving a selection of one of the plurality of simulated user models, to obtain the simulated user model, wherein each simulated user model is trained to uniquely predict a next action of the at least one simulated user in response to a simulated playback input received from the at least one simulated user.” The claim claims the non-transitory computer-readable medium counterparts corresponding to the system claim 4 and 11 above, and they are rejected, at least for the same reasons.
Regarding Claim 18, Ie discloses
“The non-transitory computer-readable medium of claim 15, further having stored thereon a sequence of instructions for causing the one or more processors to perform: generating the simulated user model by applying any one of (i) a plurality of predefined attributes of the one or more simulated users, (ii) a plurality of predefined attributes of media content items, or (iii) a combination of (i) and (ii), to the selected simulated user model.” The claim recites the non-transitory computer-readable medium counterparts corresponding to the systems of claim 6 and 12 above, and is rejected, at least for the same reasons.
Claims 2-3, 5, 10 and 16 are being rejected under pre-AIA 35 U.S.C. 103(a) as being unpatentable over Ie as applied to claim 1, 9, 15 above, and further in view of Panda (US 11360835) hereinafter referred to as Panda.
Regarding Claim 2, Ie discloses
“The system according to claim 1,”.
Ie does not disclose, “further comprising: a candidate content selection simulator configured to register a plurality of candidate content selection machine learning models, wherein the candidate content selection machine learning model is obtained by a selection of one of the plurality of candidate content selection machine learning models, wherein each of the plurality of candidate content machine learning models is trained to uniquely predict one or more selectable content media items for at least one simulated user.”
Panda further discloses, “further comprising: a candidate content selection simulator configured to register a plurality of candidate content selection machine learning models, wherein the candidate content selection machine learning model is obtained by a selection of one of the plurality of candidate content selection machine learning models, wherein each of the plurality of candidate content machine learning models is trained to uniquely predict one or more selectable content media items for at least one simulated user.” (See Detailed Description, Background and Figure 2 and 3) Panda, an analogous reference in the same field of recommender model evaluation and selection, teaches through its system that formally stores and maintains a plurality of recommender models in a structured accessible registry, systematically evaluates each model’s unique trained capabilities. And automatically selects one model from the registered plurality based on comparative performance evaluation. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to combine Ie’s simulation-based candidate agent evaluation framework with Panda’s formal model registration, characterization and selection pipeline to produce a more structured and production-ready system for registering, evaluating and selecting among candidate recommender models. (See Ie Section 1.4 and Panda Background)
Regarding Claim 3, Ie discloses
“The system of claim 2,”
Ie does not disclose, “further comprising: an input/output interface configured to receive a selection of one of the plurality of candidate content selection machine learning models.”
Panda further discloses “further comprising: an input/output interface configured to receive a selection of one of the plurality of candidate content selection machine learning models.” (See Detailed Description Col 3-4 and Fig 2, 208, wherein a system including communication or input/output interfaces configured to receive use input is disclosed). The received input is processed to select one recommender model from a plurality of recommender models based on the user input. Therefore, the input/output interface receives a selection of one of the pluralities of candidate content selection machine learning models. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to incorporate Panda’s formal input/output interface into Ie’s simulation framework as Ie acknowledges the need for a formal input/output interface through which operators drive model selection is a necessary and expected component of the simulation workflow. (See Section 4.1).
Regarding Claim 5, Ie discloses
“The system of claim 4,”
Ie does not disclose, “further comprising: an input/output interface configured to receive a selection of one of the plurality of simulated user models.”
Panda further discloses “further comprising: an input/output interface configured to receive a selection of one of the plurality of simulated user models.” (See Detailed Description Col 3-4, wherein a system including communication or input/output interfaces configured to receive use input is disclosed). The received input is processed to select one model from a plurality of models stored in memory. Therefore, the input/output interface receives a selection of one of the pluralities of simulated user models. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to incorporate Panda’s formal input/output interface into Ie’s simulation framework as Ie acknowledges that selecting among a plurality of simulated user models to accurately represent diverse user populations is a recognized requirement of the simulation workflow. (See Section 4.1).
Regarding Claim 10, Ie discloses
“The method according to claim 9,”
Ie does not disclose, “further comprising: registering a plurality of candidate content selection machine learning models; and receiving a selection of one of a plurality of candidate content selection machine learning models to obtain the candidate content selection machine learning model, wherein each of the plurality of candidate content machine learning models is trained to uniquely predict one or more selectable content media items for at least one simulated user.”
Panda further discloses “further comprising: registering a plurality of candidate content selection machine learning models; and receiving a selection of one of a plurality of candidate content selection machine learning models to obtain the candidate content selection machine learning model, wherein each of the plurality of candidate content machine learning models is trained to uniquely predict one or more selectable content media items for at least one simulated user.” (See Detailed Description Col 3-4 and Col 8, Lines 2-6 along with Fig 3, 308, wherein maintaining and selecting one model from a plurality of recommender models based on received input is disclosed. Establishes that one candidate model is obtained by selection from among the registered plurality based on systematic evaluation. Panda teaches training a plurality of recommendation models with different capabilities, resulting in distinct prediction behaviors.) It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to combine Ie’s simulation-based candidate model evaluation framework with Panda’s formal model registration and selection steps to produce a more structured and reliable method for registering and selecting among candidate content selection ML models. As Ie identifies that the comparison and selection among multiple candidate models is a core objective of its platform. (See Section 1.4).
Regarding Claim 16, Ie discloses
“The non-transitory computer-readable medium of claim 15,”
Ie does not disclose, “further having stored thereon a sequence of instructions for causing the one or more processors to perform: registering a plurality of candidate content selection machine learning models; and receiving a selection of one of a plurality of candidate content selection machine learning models to obtain the candidate content selection machine learning model, wherein each of the plurality of candidate content machine learning models is trained to uniquely predict one or more selectable content media items for at least one simulated user.”
Panda further discloses “further having stored thereon a sequence of instructions for causing the one or more processors to perform: registering a plurality of candidate content selection machine learning models; and receiving a selection of one of a plurality of candidate content selection machine learning models to obtain the candidate content selection machine learning model, wherein each of the plurality of candidate content machine learning models is trained to uniquely predict one or more selectable content media items for at least one simulated user.” (See Detailed Description Col 3-4 and Col 8, Lines 2-6, along with Fig 3, 308 wherein maintaining and selecting one model from a plurality of recommender models based on received input is disclosed. Establishes that one candidate model is obtained by selection from among the registered plurality based on systematic evaluation. Panda teaches training a plurality of recommendation models with different capabilities, resulting in distinct prediction behaviors.) It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to combine Ie’ simulation-based candidate model evaluation framework with Panda’s formal model registration and selection steps to produce a more structured and reliable method for registering and selecting among candidate content selection ML models. As Ie identifies that the comparison and selection among multiple candidate models as a core objective of its platform. (See Section 1.4).
Claims 7, 13, and 19 are being rejected under 35 U.S.C. 103 as being unpatentable over Ie and in further in view of Moore et al US 20140130076 A1 hereinafter referred to as Moore.
Regarding Claim 7, Ie discloses
“The system of claim 1, the session initializer configured to:”
Ie doesn’t disclose, “receive interaction data associated with a plurality of non-simulated users and a plurality of production initialization parameters, and initialize the content selection object by applying the plurality of session initialization parameter values to the plurality of production initialization parameters, thereby generating the initialized content selection object.”
Moore further discloses “receive interaction data associated with a plurality of non-simulated users and a plurality of production initialization parameters, and initialize the content selection object by applying the plurality of session initialization parameter values to the plurality of production initialization parameters, thereby generating the initialized content selection object.” Moore, an analogous reference in the same field of media content selection using ML-based recommendation systems, discloses, (See paragraphs [0002], [0199], [0007]. [0200], and [0197], wherein Moore discloses explicitly collecting and processing real viewer behavioral data from actual non-simulated users to initialize and configure its content selecting system. Establishes that real non-simulated user interaction data is received and used to configured the content selection system, and that interaction data from plurality of real non-simulated users is received and stored as production initialization parameters. The content selection process uses multiple production initialization parameters including viewer characteristics, location, weather, time of day, and other contextual parameters. The probability engine is initialized by applying real-world interaction data and production parameters to configure the content selection model. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to apply Moore’s production data initialization approach within Ie’s simulation framework as Ie identifies the use of real production data to initialize simulated sessions as a recognized and future development objective of its platform. (See Section 6.)
Regarding Claim 13, Ie discloses
“The method of claim 9,”
Ie doesn’t disclose, “receive interaction data associated with a plurality of non-simulated users and a plurality of production initialization parameters, and initialize the content selection object by applying the plurality of session initialization parameter values to the plurality of production initialization parameters, thereby generating the initialized content selection object.”
Moore further discloses “receive interaction data associated with a plurality of non-simulated users and a plurality of production initialization parameters, and initialize the content selection object by applying the plurality of session initialization parameter values to the plurality of production initialization parameters, thereby generating the initialized content selection object.” Moore, an analogous reference in the same field of media content selection using ML-based recommendation systems, discloses, (See paragraphs [0002], [0199], [0007]. [0200], and [0197], wherein Moore discloses explicitly collecting and processing real viewer behavioral data from actual non-simulated users to initialize and configure its content selecting system. Establishes that real non-simulated user interaction data is received and used to configured the content selection system, and that interaction data from plurality of real non-simulated users is received and stored as production initialization parameters. The content selection process uses multiple production initialization parameters including viewer characteristics, location, weather, time of day, and other contextual parameters. The probability engine is initialized by applying real-world interaction data and production parameters to configure the content selection model. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to apply Moore’s production data initialization approach within Ie’s simulation framework as Ie identifies the use of real production data to initialize simulated sessions as a recognized and future development objective of its platform. (See Section 6.)
Regarding Claim 19, Ie discloses
“The non-transitory computer-readable medium of claim 15, further having stored thereon a sequence of instructions for causing the one or more processors to perform:”
Ie doesn’t disclose, “receive interaction data associated with a plurality of non-simulated users and a plurality of production initialization parameters, and initialize the content selection object by applying the plurality of session initialization parameter values to the plurality of production initialization parameters, thereby generating the initialized content selection object.”
Moore further discloses “receive interaction data associated with a plurality of non-simulated users and a plurality of production initialization parameters, and initialize the content selection object by applying the plurality of session initialization parameter values to the plurality of production initialization parameters, thereby generating the initialized content selection object.” Moore, an analogous reference in the same field of media content selection using ML-based recommendation systems, discloses, (See paragraphs [0002], [0199], [0007]. [0200], and [0197], wherein Moore discloses explicitly collecting and processing real viewer behavioral data from actual non-simulated users to initialize and configure its content selecting system. Establishes that real non-simulated user interaction data is received and used to configured the content selection system, and that interaction data from plurality of real non-simulated users is received and stored as production initialization parameters. The content selection process uses multiple production initialization parameters including viewer characteristics, location, weather, time of day, and other contextual parameters. The probability engine is initialized by applying real-world interaction data and production parameters to configure the content selection model. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to apply Moore’s production data initialization approach within Ie’s simulation framework as Ie identifies the use of real production data to initialize simulated sessions as a recognized and future development objective of its platform. (See Section 6.)
Claims 8, 14, 20 are being rejected under pre-AIA 35 U.S.C. 103(a) as being unpatentable over Ie as applied to claim 1, 9, 15 above, and further in view of Quadrana et al. “Personalizing Session-based Recommendations with Hierarchical Recurrent Neural Networks” hereinafter referred to as Quadrana.
Regarding Claim 8, Ie discloses
“The system of claim 1,”
Ie doesn’t disclose, “wherein the session initializer is further configured to: select a recorded session of a real user, and initialize the candidate content selection machine learning model to a particular time within the recorded session.”
Quadrana, an analogous reference in the same field of session-based recommendation systems that use recorded real user session data to initialize ML models at particular temporal points within those sessions, further discloses, “wherein the session initializer is further configured to: select a recorded session of a real user, and initialize the candidate content selection machine learning model to a particular time within the recorded session.” (See Sections 4.1 and 4.3 along with Equation 4, wherein it is discloses selecting recorded prior real user sessions for model initialization and bootstrapping purposes. Also discloses initializing the session-level GRU to a particular state corresponding to a specific time point within recorded user sessions. The initial hidden state of the session-level GRU for a forthcoming session is set to,
PNG
media_image1.png
16
157
media_image1.png
Greyscale
, where cm is the user-level representation derived from the selected prior recorded sessions, establishing that the candidate ML model is initialized to a state corresponding to a particular time point within the recorded session history. The model is initialized to correspond to the specific temporal state of the recorded session at the point where initialization occurs. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to combine the teachings of Ie in simulation-based session initialization framework with Quadrana’s technique of selecting recorded real user sessions and initializing the candidate ML model to a particular time within those sessions as Quadrana explicitly confirms, (See Section 4.3), that temporal initialization of model within a recorded session is a recognized and necessary step is session-based recommendation systems. Ie also identifies the use of real recorded user sessions for the model initialization as a recognized need and planned development of its platform. (See Section 6).
Regarding Claim 14, Ie discloses
“The method of claim 9,”
Ie doesn’t disclose, “further comprising: selecting a recorded session of a real user; and initializing the candidate content selection machine learning model to a particular time within the recorded session.”
Quadrana, an analogous reference in the same field of session-based recommendation systems that use recorded real user session data to initialize ML models at particular temporal points within those sessions, further discloses, “further comprising: selecting a recorded session of a real user; and initializing the candidate content selection machine learning model to a particular time within the recorded session.” (See Sections 4.1 and 4.3 along with Equation 4, wherein it is discloses selecting recorded prior real user sessions for model initialization and bootstrapping purposes. Also discloses initializing the session-level GRU to a particular state corresponding to a specific time point within recorded user sessions. The initial hidden state of the session-level GRU for a forthcoming session is set to,
PNG
media_image1.png
16
157
media_image1.png
Greyscale
, where cm is the user-level representation derived from the selected prior recorded sessions, establishing that the candidate ML model is initialized to a state corresponding to a particular time point within the recorded session history. The model is initialized to correspond to the specific temporal state of the recorded session at the point where initialization occurs. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to combine the teachings of Ie in simulation-based session initialization framework with Quadrana’s technique of selecting recorded real user sessions and initializing the candidate ML model to a particular time within those sessions as Quadrana explicitly confirms, (See Section 4.3), that temporal initialization of model within a recorded session is a recognized and necessary step is session-based recommendation systems. Ie also identifies the use of real recorded user sessions for the model initialization as a recognized need and planned development of its platform. (See Section 6).
Regarding Claim 20, Ie discloses
“The non-transitory computer-readable medium of claim 15,”
Ie doesn’t disclose, “further having stored thereon a sequence of instructions for causing the one or more processors to perform: selecting a recorded session of a real user; and initializing the candidate content selection machine learning model to a particular time within the recorded session.”
Quadrana, an analogous reference in the same field of session-based recommendation systems that use recorded real user session data to initialize ML models at particular temporal points within those sessions, further discloses, “further having stored thereon a sequence of instructions for causing the one or more processors to perform: selecting a recorded session of a real user; and initializing the candidate content selection machine learning model to a particular time within the recorded session.” (See Sections 4.1 and 4.3 along with Equation 4, wherein it is discloses selecting recorded prior real user sessions for model initialization and bootstrapping purposes. Also discloses initializing the session-level GRU to a particular state corresponding to a specific time point within recorded user sessions. The initial hidden state of the session-level GRU for a forthcoming session is set to,
PNG
media_image1.png
16
157
media_image1.png
Greyscale
, where cm is the user-level representation derived from the selected prior recorded sessions, establishing that the candidate ML model is initialized to a state corresponding to a particular time point within the recorded session history. The model is initialized to correspond to the specific temporal state of the recorded session at the point where initialization occurs. It would have been obvious to a person having ordinary skill in the art before the effective filling date of the claimed invention to combine the teachings of Ie in simulation-based session initialization framework with Quadrana’s technique of selecting recorded real user sessions and initializing the candidate ML model to a particular time within those sessions as Quadrana explicitly confirms, (See Section 4.3), that temporal initialization of model within a recorded session is a recognized and necessary step is session-based recommendation systems. Ie also identifies the use of real recorded user sessions for the model initialization as a recognized need and planned development of its platform. (See Section 6).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MEHNAZ J. SIDDIQUEE whose telephone number is (571)272-1366. The examiner can normally be reached M-F.
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, Renee Chavez can be reached at (571) 270-1104. 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.
/MEHNAZ JAREEN SIDDIQUEE/Examiner, Art Unit 2186
/RENEE D CHAVEZ/Supervisory Patent Examiner, Art Unit 2186
Prosecution Insights