DETAILED ACTION
This communication is a Final Office Action rejection on the merits. Claims 1-20 are currently pending and have been addressed below.
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 .
Response to Arguments
Applicant's arguments filed 08/07/2026 (related to the 103 Rejection) have been fully considered but are moot in view of new grounds of rejection. Applicant's amendments necessitated the new ground(s) of rejection presented in this Office action. Rejection based on a newly cited reference(s) follows.
Applicant's arguments filed 08/07/2026 (related to the 101 Rejection) have been fully considered but they are not persuasive.
Applicant states, on pages 14-17, that the claims do not fall into any of the abstract ideas exceptions provided by the Guidance.
Examiner respectfully disagrees with Applicant. These claim elements are considered to be abstract ideas because they are directed to “commercial or legal interactions” and “mathematical calculations.” In this case, “distributing content to one or more target users based on predicted demographic/interest information” is a form of advertisement. Also, the new limitations of “training …, linking data …, embedding …, normalizing …, and predicting multiple pieces of demographic/interest information of the user” are merely mathematical calculations. If a claim limitation, under its broadest reasonable interpretation, covers commercial interactions or mathematical calculations, then it falls within the “method of organizing human activity” and/or “mathematical concepts” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
Applicant further states, on pages 17-22, that the claims provide an improvement to known technical problems of traditional individual prediction of demographic information, such as high requirement in resource, inefficient operation, and failure to consider interplay among data. (Paras. [0002]-[0005]). Applicant's claimed concept overcomes such technical problems by "jointly predicting demographic/interest information based on data from different sources via joint modeling and content targeting using such jointly predicted information. Instead of individually predicting different pieces of information, which is not only inefficient but also does not consider the interactions among different data from different platforms/sources, the framework as disclosed herein models the interplay of different pieces and types of data gathered from different platforms for jointly predicting different pieces of demographic/interest information for targeting." (Para. [0030]).
Also, the claims are necessarily rooted in computer technology of machine learning and provide a solution to "model[] the interplay of different pieces and types of data gathered from different platforms for jointly predicting different pieces of demographic/interest information for targeting." (Para. [0030]).
The recited features are clearly tied to a practical application, i.e., training a joint prediction model by machine learning via a perceptron neural network and utilizing the trained model to make demographics/interests prediction based on normalized input data from multiple different platforms/sources.
Examiner respectfully disagrees with Applicant. The main functions of the additional elements of “processor” and “perceptron neural network” are merely used to: collect data (e.g., data form different sources on different platforms), analyze the data (e.g., link data from different sources, embed the data, normalize the data, and predict multiple pieces of demographic/interest information), and display certain results of the collection and analysis (e.g., output predictions). Those are functions that the courts have described as merely indicating a field of use or technological environment in which to apply a judicial exception (see MPEP 2106.05f). In this case, although the claim recites “training based on data streams representing trails of a plurality of users as dynamic input via a perceptron neural network,” the claim does not provide any details about how the trained perceptron neural network operates and/or how the embeddings are generated (e.g., the perceptron neural network is merely used to analyze input data from different sources and output prediction of multiple pieces of demographic or interest information). Thus, the perceptron neural network is a black box, which is merely claiming the idea of a solution or outcome (MPEP 2106.05a). Also, the claim and specification do not provide any language stating how the perceptron neural network is improved upon prior art. The claim further states “distributing content to one or more target users selected based on the multiple pieces of demographic or interest information associated with each of the one or more target users.” In this case, those limitations are considered an “insignificant extra-solution activity” since adding a final step of “distributing content based on the generated information” does not add a meaningful limitation to the process of predicting demographic/interest information (see MPEP 2106.05(g)).
Lastly, the claim fails to recite any improvements to another technology or technical field, improvements to the functioning of the computer itself, use of a particular machine, effecting a transformation or reduction of a particular article to a different state or thing, adding unconventional steps that confine the claim to a particular useful application, and/or meaningful limitations beyond generally linking the use of an abstract idea to a particular environment. See 84 Fed. Reg. 55. Viewed individually or as a whole, these additional claim element(s) do not provide meaningful limitation(s) to transform the abstract idea into a patent eligible application of the abstract idea such that the claim(s) amounts to significantly more than the abstract idea itself. Thus, the claim is not patent eligible.
Independent claims 8 and 15 recite similar features and therefore are rejected for the same reasons as independent claim 1. Claims 2-7, 9-14, and 16-20 are rejected for having the same deficiencies as those set forth with respect to the claims that they depend from, independent claims 1, 8, and 15.
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 a judicial exception (i.e., an abstract idea) without reciting significantly more.
Independent Claim 1
Step One - First, pursuant to step 1 in the January 2019 Revised Patent Subject Matter Eligibility Guidance (“2019 PEG”) on 84 Fed. Reg. 53, the claim 1 is directed to a method which is a statutory category.
Step 2A, Prong One - Claim 1 recites: A method for prediction of demographics or interests, comprising: training, based on data streams representing trails of a plurality of users as dynamic input, a joint prediction model by dynamically learning based on the dynamic input, wherein the data streams are from different sources on different platforms providing data with respect to different types of identifications and different types of events; receiving data from different sources (DFDS) having identifications included therein, wherein the DFDS relates to a plurality of users identifiable via the identifications; linking DFDS associated with each of the plurality of users based on the identifications so that each group of the linked DFDS is under a same identification; for each group of the linked DFDS associated with each of the plurality of users, training, based on a portion of the group of the linked DFDS, representing trails of the user; normalizing the linked DFDS collected from the different platforms, wherein ratings from the plurality of users are rescaled so that all rating related data are recorded using a uniform scale without changing relative evaluations from different users; generating an input vector for the user based on the group of the normalized DFDS including the trained, by using the trained to map the portion of the group of the linked DFDS to a sub-vector of the input vector, wherein the portion has a higher dimension and the sub-vector has a lower dimension; and generating an output vector by predicting, using the trained joint prediction model based on the input vector associated with the user, multiple pieces of demographic or interest information of the user, wherein the input vector has a higher dimensionality than the output vector, and wherein attributes included in the output vector are the predicted multiple pieces of demographic or interest information, wherein the predicted multiple pieces comprise a gender, an age, a residence region, and a profession; and distributing content to one or more target users selected based on the multiple pieces of demographic or interest information associated with each of the one or more target users. These claim elements are considered to be abstract ideas because they are directed to “commercial or legal interactions” and “mathematical calculations.” In this case, “distributing content to one or more target users based on predicted demographic/interest information” is a form of advertisement. Also, the new limitations of “training …, linking data …; embedding …, normalizing …, and predicting multiple pieces of demographic or interest information of the user” are merely mathematical calculations. If a claim limitation, under its broadest reasonable interpretation, covers commercial interactions or mathematical calculations, then it falls within the “method of organizing human activity” and/or “mathematical concepts” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
Step 2A Prong 2 - The judicial exception is not integrated into a practical application. Claim 1 includes additional elements: a processor; a memory; a communication platform; and training via a perceptron neural network; a joint prediction model; and embeddings
The processor is merely used to execute instructions and receive data from different sources (Paragraphs 0044 & 0060). The memory is merely used to store information (Paragraph 0061). The communication platform is merely used to communicate data (Paragraph 0058). The multiple layer perception neural network is merely used to implement a joint demographic/interest prediction (Paragraphs 0026 & 0052). The joint prediction model is merely used to predict multiple pieces of demographic or interest information of the user (Paragraph 0007). The “embeddings” correspond to parameters of a mechanism that, once trained, may be used to generate an output, representing, e.g., a feature vector. That is, the mechanism with trained embeddings may take data in a certain dimension as input and then generate an output vector with attributes characterizing the input in a different dimension. Such embeddings may be derived via training and the trained embeddings may be provided as a service (Paragraph 0038). Merely stating that the step is performed by a computer component results in “apply it” on a computer (MPEP 2106.05f). These elements of “processor,” “memory,” “communication platform,” “perceptron neural network,” “joint prediction model,” and “embeddings” are recited at a high level of generality such that it amounts no more than mere instructions to apply the exception using a generic computer element. In this case, although the claim recites “training embeddings representing trails of the user,” the claim does not provide any details about how the trained perceptron neural network operates or how the embeddings are generated, which amounts to no more than mere instructions to apply the exception using a generic computer (MPEP 2106.05(f)). Accordingly, alone and in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea. Therefore, the claim is directed to an abstract idea.
Step 2B - The claim does not include additional elements that are sufficient to amount significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the claims describe how to generally “apply” the concept of: predicting multiple pieces of demographic or interest information of each of the plurality of users; and distributing content to the users based on the predicted demographic/interest information. The specification shows that the processor is merely used to execute instructions and receive data from different sources (Paragraphs 0044 & 0060). The memory is merely used to store information (Paragraph 0061). The communication platform is merely used to communicate data (Paragraph 0058). The multiple layer perception neural network is merely used to implement a joint demographic or interest prediction (Paragraphs 0026 & 0052). The joint prediction model is merely used to predict multiple pieces of demographic/interest information of the user (Paragraph 0007). The “embeddings” correspond to parameters of a mechanism that, once trained, may be used to generate an output, representing, e.g., a feature vector. That is, the mechanism with trained embeddings may take data in a certain dimension as input and then generate an output vector with attributes characterizing the input in a different dimension. Such embeddings may be derived via training, and the trained embeddings may be provided as a service (Paragraph 0038). Also, the claim and specification do not provide any language stating how the perceptron neural network is improved upon prior art (MPEP 2106.05a). Thus, nothing in the claim adds significantly more to the abstract idea. The claim is ineligible.
Independent claim 8 is directed to an article of manufacture at step 1, which is a statutory category. Claim 8 recites similar limitations as claim 1 and is rejected for the same reasons at step 2a, prong one; step 2a, prong 2; and step 2b. Claim 8 further recites “non-transitory medium” and “machine” – which are treated as just an explicit “computer component” for executing the operations (MPEP 2106.05f). Accordingly, these additional elements are viewed as “apply it on a computer” at step 2a, prong 2 and step 2B. Thus, nothing in the claim adds significantly more to the abstract idea. The claim is ineligible.
Independent claim 15 is directed to an apparatus at step 1, which is a statutory category. Claim 15 recites similar limitations as claim 1 and is rejected for the same reasons at step 2a, prong one; step 2a, prong 2; and step 2b. Claim 15 does not recite any additional elements to consider under step 2a, prong 2 and step 2B. Thus, nothing in the claim adds significantly more to the abstract idea. The claim is ineligible.
Dependent claims 2, 4-7, 9, 11-14, 16, and 18-20 are not directed to any additional claim elements. Rather, these claims offer further descriptive limitations of the abstract idea mentioned above - such as: describing the plurality of sources and identification obtained from the plurality of sources; wherein the predictions are performed simultaneously; specifying the multiple pieces of demographic/interest information such as age, gender, residence, and profession; and a step for selecting the one or more target users for providing content. These processes are similar to the abstract idea noted in the independent claim because they further the limitations of the independent claim which are directed to “certain methods of organizing human activity” which include “commercial or legal interactions.” In addition, no additional elements are integrated into the abstract idea. Therefore, the claims still recite an abstract idea that can be grouped into certain methods of organizing human activity.
Dependent claims 3, 10, and 17 are directed to an additional element such as: application graph data. The application graph data is merely used to capture the dynamics of applications' installation, usage level, activities therein, peak usage time, valley usage time, correlation with other application usage patterns, etc. may be established and used as part of the DFDS input to the joint model (Paragraphs 0042). The application graph data is recited at a high level of generality such that it amounts no more than mere instructions to apply the exception using a generic computer element (MPEP 2106.05f). Thus, nothing in the claim adds significantly more to the abstract idea. The claim is ineligible.
Examiner recommends to follow Example 47 of the 2024 AI Guidance Update, if supported by the specification.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-6, 8-13, 15-19 are rejected under 35 U.S.C. 103 as being unpatentable over Podoynitsina (Podoynitsina, L., Romanenko, A., Kryzhanovskiy, K. and Moiseenko, A., 2017. Demographic prediction based on mobile user data. Electronic Imaging, 29, pp.44-47), in view of Malmi et al. (US 2018/0189660 A1), in further view of Zatorski et al. (US 2023/0100788 A1).
Regarding claim 1 (Currently Amended), Podoynitsina discloses a method implemented on at least one processor, a memory, and a communication platform for prediction of demographics or interests, comprising (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age);
training, based on data streams representing trails of a plurality of users as dynamic input via a perceptron neural network, a joint prediction model by dynamically learning based on the dynamic input, wherein the data streams are from different sources on different platforms providing data with respect to different types of identifications and different types of events; receiving data from different sources (DFDS) having identifications included therein, wherein the DFDS relates to a plurality of users identifiable via the identifications (Page 45, Demographic Model, Once the information about topics is aggregated, we need to build a demographic model. Demographic model consist of several demographic classifies. In our case: age classifier, gender classifier and marital status classifier. In this work, we chose to use the deep learning approach using the Veles framework. Each classifier was built with neural network and optimized with genetic algorithm. The architecture of neural network is based on the multi-layer perceptron (Collobert and Bengio, 2004[18]); Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android platform, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data. Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms);
linking DFDS associated with each of the plurality of users based on the identifications so that each group of the linked DFDS is under a same identification (Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android plat form, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data. Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms; Examiner notes that DFDS (e.g., web page data and application data) is associated to the same user (e.g., mobile device));
for each group of the linked DFDS associated with each of the plurality of users, training, based on a portion of the group of the linked DFDS, embeddings representing trails of the user (Page 44, Introduction, First we should build and train topic model with classifier of text data to specified categories (interests). The list of wanted categories can be given by content provider. The topic model categorizes the text extracted from Web pages. We build topic model with the Additive Regularization of Topic Models (ARTM) algorithm. Then we extract user interests using trained topic model. ARTM can be used not only for clustering, but for classification for a given list of categories. ARTM is based on generalization of two powerful algorithms: probabilistic latent semantic analysis (PLSA) [9] and latent Dirichlet allocation (LDA); Page 45, Document Aggregation, The resulting topic vector (or user interest vector) is used as feature vector for demographic model; Page 46, Demographic Model, We used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model; Page 46, Through the early tests, we chose the NN approach to build demographic prediction classifier. Currently different deep learning frameworks are available for training NNs: Caffe [13], Torch [14], Theano [15] and others. However, we used our custom deep learning framework (Veles) [16] because it was designed as a very flexible tool in terms of workflow construction, data extraction and preprocessing, visualization, with an additional advantage in the ease of porting of the resulting classifier to mobile devices; As stated in Paragraph 0038 of Applicant’s specification, embeddings may take data in a certain dimension as input and then generate an output vector with attributes characterizing the input in a different dimension. Therefore, based on broadest reasonable interpretation in light of the specification, Podoynitsina discloses “embeddings representing trails of the user” since it can reduce the size of an input feature vector. In this case, the input feature vector is the text extracted from webpages);
normalizing the linked DFDS collected from the different platforms, …; generating an input vector for the user based on the group of the normalized DFDS including the trained embeddings, by using the trained embeddings to map the portion of the group of the linked DFDS to a sub-vector of the input vector, wherein the portion has a higher dimension and the sub-vector has a lower dimension (Page 44, Introduction, First we should build and train topic model with classifier of text data to specified categories (interests). The list of wanted categories can be given by content provider. The topic model categorizes the text extracted from Web pages. We build topic model with the Additive Regularization of Topic Models (ARTM) algorithm. Then we extract user interests using trained topic model; Page 45, Preprocessing, The major source of our observation data is webpages browsed by the user. We preprocess the web pages as follows: remove HTML tags, perform stemming or lemmatization of every word, remove stop words, lowercase all characters and translate webpage content into a target languages; Page 45, Document Aggregation, The resulting topic vector (or user interest vector) is used as feature vector for demographic model; Page 46, Demographic Model, We used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model; Examiner interprets “preprocessing the data” as “normalizing the data”);
and generating an output vector by predicting, using the trained joint prediction model based on the input vector associated with the user, multiple pieces of demographic or interest information of the user (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age), wherein the input vector has a higher dimensionality than the output vector (Page 46, Demographic Model, We use several hyper-parameters of the neural network architecture: size of minibatch, number of layers, number of neurons in each layer, activation function, dropout, learning rate, weights de cay, gradient moment, standard deviation of weights, gradient de scent step, regularization coefficients, initial ranges of weights, number of examples per iteration. Genetic algorithm enables us to adjust these hyper-parameters. Also, we used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model), and wherein attributes included in the output vector are the predicted multiple pieces of demographic or interest information, wherein the predicted multiple pieces comprise a gender, an age, … (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age);
and distributing content to one or more target users selected based on the multiple pieces of demographic or interest information associated with each of the one or more target users (Page 44, Introduction, Our task is to build a demographical model, which will recognize demographic characteristics of user, such as gender, marital status and age. Such demographic information play a crucial role in personalized services and targeted advertising).
Although Podoynitsina discloses predicting multiple pieces of demographic or interest information (e.g., gender and age), Podoynitsina does not specifically disclose wherein the multiple pieces of demographic information further includes a residence region and a profession.
However, Malmi et al. discloses a method implemented on at least one processor, a memory, and a communication platform for prediction of demographics or interests, comprising (Paragraph 0012, In one aspect, there is provided a method for predicting user demographics, and therewith e.g. user preferences, interests, purchase intent and/or other behavior or characteristics in view of providing personalized digital content, to be performed by at least one electronic apparatus, optionally a number of functionally connected servers potentially at least some which being operable in a cloud computing environment; Paragraph 0084, The terminal devices 104a, 104b, 104c, 104d, 104e, 104f and/or external devices/systems 114, 115, 116 directly or indirectly connected to the arrangement 114 for providing data thereto or obtaining data such as various deliverables therefrom, may generally contain similar hardware elements such as a processor, a memory and a communication interface): …
and wherein attributes included in the output vector are the predicted multiple pieces of demographic or interest information, wherein the predicted multiple pieces comprise a gender, an age, a residence region, and a profession (Paragraph 0041, The arrangement may be configured to contain or implement a number of functional modules to execute different method items, such as a data collector module for obtaining deterministic user data for modeling, a modeler for actually creating the model(s), a metering module for capturing (deterministic) usage statistics of apps for prediction purposes, and a predictor to estimate (predict) demographics based on the metered app usage data; Paragraph 0086, Model data collector 312 may be configured to receive (optionally by interrogation/pulling/survey mechanism) and manage (store, filter, combine, process, distribute) user data for modeling purposes. The data may contain hard, deterministic data having regard to a number of demographic characteristics (e.g. age, sex, marital status, race, income, language, country or other location information, occupation, and/or religion), application usage statistics and optional further, potentially behavioral data, which may be then utilized in creating a number of models associating e.g. app usage information (explanatory variables) with related predicted demographic characteristics (dependent variables) as explained herein. Further, a number of other characteristics such as behavioral characteristics may be included in the model as explanatory and/or dependent variables).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information (e.g., age and gender) of the invention of Podoynitsina to further specify wherein the multiple pieces of demographic information further includes a residence region and a profession of the invention of Malmi et al. because doing so would allow the method to predict other demographic information such as age, location information, and occupation (see Malmi et al., Paragraph 0041). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Although Podoynitsina discloses normalizing the DFDS collected from the different platforms (e.g., preprocessing data obtained from the app store and webpages), Podoynitsina does not specifically disclose wherein the normalization of data includes rescaling rating related data to a uniform scale.
However, Zatorski et al. discloses for each group of the linked DFDS associated with each of the plurality of users, training, based on a portion of the group of the linked DFDS, embeddings representing trails of the user (Paragraph 0007, According to an embodiment of the disclosure, a method of recommending content to a user may include based on a content recommendation request received from a device of the user. A Method may include obtaining a feedback vector of the user for at least one of a plurality of pieces of content. A Method may include obtaining feedback information for the plurality of pieces of content used by a plurality of users. A Method may include generating user embedding vectors for the plurality of users and content embedding vectors for the plurality of pieces of content, based on the feedback information. A Method may include determining a plurality of user groups and central users of the plurality of user groups by grouping the plurality of users based on the user embedding vectors for the plurality of users. A Method may include inputting user embedding vectors for the central users and the content embedding vectors for the plurality of pieces of content to an artificial neural network model configured to estimate a preference degree for content. A Method may include obtaining preference vectors of the central users, which indicate estimated preference degrees of the central users for the plurality of pieces of content, the preference vectors being output by the artificial neural network model. A Method may include determining at least one piece of recommendation content to be recommended to the user, based on the feedback vector of the user and the preference vectors of the central users. A Method may include recommending the determined at least one piece of recommendation content to the user; Paragraph 0046, The user embedding vectors 1020 may include a user embedding vector 1021 of the first user 1001, a user embedding vector 1022 of the second user 1002, a user embedding vector 1023 of the third user 1003, a user embedding vector 1024 of a fourth user (not shown), a user embedding vector 1025 of a fifth user (not shown), and a user embedding vector 1026 of a sixth user (not shown), but are not limited thereto. The content embedding vectors 1030 may include a content embedding vector 1031 of first content, a content embedding vector 1032 of second content, a content embedding vector 1033 of third content, a content embedding vector 1034 of fourth content, and a content embedding vector 1035 of fifth content, but are not limited thereto; Paragraph 0047, Also, the server 1000 may train an artificial neural network model based on the user embedding vectors 1020 of the plurality of users and the content embedding vectors 1030 of the plurality of pieces of content in operation 1040. The artificial neural network model 1070 may be used to estimate a preference degree of a user for content; Paragraph 0188, The device 3200 may include a smartphone, a tablet personal computer (PC), a PC, a smart television (TV), a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, an electronic book terminal, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, or another mobile or non-mobile computing device, but is not limited thereto);
normalizing the linked DFDS collected from the different platforms, wherein ratings from the plurality of users are rescaled so that all rating related data are recorded using a uniform scale without changing relative evaluations from different users (Paragraph 0035, In the present specification, content may include multimedia content including movies, music, and the like, and content including various products purchasable from the Internet or an application downloadable from an application store; Paragraph 0074, According to an embodiment of the disclosure, a feedback vector of a user may be a vector indicating a preference degree of the user for at least one of a plurality of pieces of content. In detail, the preference degree of the user for at least one of the plurality of pieces of content may be determined based on a behavior or reaction of the user for the at least one of the plurality of pieces of content, and the preference degrees of the user for the plurality of pieces of content may be classified as “positive” or “negative”, or “positive”, “negative”, or “no preference”, but are not limited thereto. For example, the preference degrees of the user for the plurality of pieces of content may be classified as “positive” or “negative”, or “positive”, “negative”, or “no preference”, based on behaviors of the user clicking the plurality of pieces of content, scrolling the plurality of pieces of content, looking at the plurality of pieces of content for at least a certain period of time, scoring the plurality of pieces of content, downloading the plurality of pieces of content, streaming the plurality of pieces of content for at least a certain period of time, or watching previews of the plurality of pieces of content. Also, for example, a method of classifying the preference degrees of the user for the plurality of pieces of content as “positive” or “negative”, or “positive”, “negative”, or “no preference” may be stored in the server after being changed or pre-set by a processor of the server; Paragraph 0188, The device 3200 may include a smartphone, a tablet personal computer (PC), a PC, a smart television (TV), a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, an electronic book terminal, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, or another mobile or non-mobile computing device, but is not limited thereto; Examiner interprets “rating the plurality of pieces of content as positive or negative” as “normalizing the data to a uniform scale”);
generating an input vector for the user based on the group of the normalized DFDS including the trained embeddings, by using the trained embeddings to map the portion of the group of the linked DFDS to a sub-vector of the input vector, wherein the portion has a higher dimension and the sub-vector has a lower dimension (Paragraph 0096, The server may change the high-dimensional user embedding vectors for the plurality of users to the pre-set low-dimensional user embedding vectors for the plurality of users, through a principal component analysis (PCA)).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information of each of the plurality of users based on data received from multiple sources of the invention of Podoynitsina to further specify how the data received from multiple sources is embedded and normalized of the invention of Zatorski et al. because doing so would allow the method to input user embedding vectors for the central users and the content embedding vectors for the plurality of pieces of content to an artificial neural network model configured to estimate a preference degree for content (see Zatorski et al., Paragraph 0007). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Regarding claim 8 (Currently Amended), Podoynitsina discloses a [method/system] having information recorded thereon for prediction of demographics or interests, wherein the information, when read by the machine, causes the machine to perform the following steps (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age);
training, based on data streams representing trails of a plurality of users as dynamic input via a perceptron neural network, a joint prediction model by dynamically learning based on the dynamic input, wherein the data streams are from different sources on different platforms providing data with respect to different types of identifications and different types of events; receiving data from different sources (DFDS) having identifications included therein, wherein the DFDS relates to a plurality of users identifiable via the identifications (Page 45, Demographic Model, Once the information about topics is aggregated, we need to build a demographic model. Demographic model consist of several demographic classifies. In our case: age classifier, gender classifier and marital status classifier. In this work, we chose to use the deep learning approach using the Veles framework. Each classifier was built with neural network and optimized with genetic algorithm. The architecture of neural network is based on the multi-layer perceptron (Collobert and Bengio, 2004[18]); Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android plat form, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data. Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms);
linking DFDS associated with each of the plurality of users based on the identifications so that each group of the linked DFDS is under a same identification (Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android plat form, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data. Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms; Examiner notes that DFDS (e.g., web page data and application data) is associated to the same user (e.g., mobile device));
for each group of the linked DFDS associated with each of the plurality of users, training, based on a portion of the group of the linked DFDS, embeddings representing trails of the user (Page 44, Introduction, First we should build and train topic model with classifier of text data to specified categories (interests). The list of wanted categories can be given by content provider. The topic model categorizes the text extracted from Web pages. We build topic model with the Additive Regularization of Topic Models (ARTM) algorithm. Then we extract user interests using trained topic model. ARTM can be used not only for clustering, but for classification for a given list of categories. ARTM is based on generalization of two powerful algorithms: probabilistic latent semantic analysis (PLSA) [9] and latent Dirichlet allocation (LDA); Page 45, Document Aggregation, The resulting topic vector (or user interest vector) is used as feature vector for demographic model; Page 46, Demographic Model, We used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model; Page 46, Through the early tests, we chose the NN approach to build demographic prediction classifier. Currently different deep learning frameworks are available for training NNs: Caffe [13], Torch [14], Theano [15] and others. However, we used our custom deep learning framework (Veles) [16] because it was designed as a very flexible tool in terms of workflow construction, data extraction and preprocessing, visualization, with an additional advantage in the ease of porting of the resulting classifier to mobile devices; As stated in Paragraph 0038 of Applicant’s specification, embeddings may take data in a certain dimension as input and then generate an output vector with attributes characterizing the input in a different dimension. Therefore, based on broadest reasonable interpretation in light of the specification, Podoynitsina discloses “embeddings representing trails of the user” since it can reduce the size of an input feature vector. In this case, the input feature vector is the text extracted from webpages);
normalizing the linked DFDS collected from the different platforms, …; generating an input vector for the user based on the group of the normalized DFDS including the trained embeddings, by using the trained embeddings to map the portion of the group of the linked DFDS to a sub-vector of the input vector, wherein the portion has a higher dimension and the sub-vector has a lower dimension (Page 44, Introduction, First we should build and train topic model with classifier of text data to specified categories (interests). The list of wanted categories can be given by content provider. The topic model categorizes the text extracted from Web pages. We build topic model with the Additive Regularization of Topic Models (ARTM) algorithm. Then we extract user interests using trained topic model; Page 45, Preprocessing, The major source of our observation data is webpages browsed by the user. We preprocess the web pages as follows: remove HTML tags, perform stemming or lemmatization of every word, remove stop words, lowercase all characters and translate webpage content into a target languages; Page 45, Document Aggregation, The resulting topic vector (or user interest vector) is used as feature vector for demographic model; Page 46, Demographic Model, We used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model; Examiner interprets “preprocessing the data” as “normalizing the data”);
and generating an output vector by predicting, using the trained joint prediction model based on the input vector associated with the user, multiple pieces of demographic or interest information of the user (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age), wherein the input vector has a higher dimensionality than the output vector (Page 46, Demographic Model, We use several hyper-parameters of the neural network architecture: size of minibatch, number of layers, number of neurons in each layer, activation function, dropout, learning rate, weights de cay, gradient moment, standard deviation of weights, gradient de scent step, regularization coefficients, initial ranges of weights, number of examples per iteration. Genetic algorithm enables us to adjust these hyper-parameters. Also, we used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model), and wherein attributes included in the output vector are the predicted multiple pieces of demographic or interest information, wherein the predicted multiple pieces comprise a gender, an age, … (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age);
and distributing content to one or more target users selected based on the multiple pieces of demographic or interest information associated with each of the one or more target users (Page 44, Introduction, Our task is to build a demographical model, which will recognize demographic characteristics of user, such as gender, marital status and age. Such demographic information play a crucial role in personalized services and targeted advertising).
Although Podoynitsina discloses predicting multiple pieces of demographic or interest information (e.g., gender and age), Podoynitsina does not specifically disclose wherein the multiple pieces of demographic information further includes a residence region and a profession.
However, Malmi et al. discloses machine readable and non-transitory medium having information recorded thereon for prediction of demographics or interests, wherein the information, when read by the machine, causes the machine to perform the following steps (Paragraph 0012, In one aspect, there is provided a method for predicting user demographics, and therewith e.g. user preferences, interests, purchase intent and/or other behavior or characteristics in view of providing personalized digital content, to be performed by at least one electronic apparatus, optionally a number of functionally connected servers potentially at least some which being operable in a cloud computing environment; Paragraph 0080, A computer program product comprising the appropriate software code means may be provided. It may be embodied in a non-transitory carrier medium such as a memory card, an optical disc or a USB (Universal Serial Bus) stick, for example): …
and wherein attributes included in the output vector are the predicted multiple pieces of demographic or interest information, wherein the predicted multiple pieces comprise a gender, an age, a residence region, and a profession (Paragraph 0041, The arrangement may be configured to contain or implement a number of functional modules to execute different method items, such as a data collector module for obtaining deterministic user data for modeling, a modeler for actually creating the model(s), a metering module for capturing (deterministic) usage statistics of apps for prediction purposes, and a predictor to estimate (predict) demographics based on the metered app usage data; Paragraph 0086, Model data collector 312 may be configured to receive (optionally by interrogation/pulling/survey mechanism) and manage (store, filter, combine, process, distribute) user data for modeling purposes. The data may contain hard, deterministic data having regard to a number of demographic characteristics (e.g. age, sex, marital status, race, income, language, country or other location information, occupation, and/or religion), application usage statistics and optional further, potentially behavioral data, which may be then utilized in creating a number of models associating e.g. app usage information (explanatory variables) with related predicted demographic characteristics (dependent variables) as explained herein. Further, a number of other characteristics such as behavioral characteristics may be included in the model as explanatory and/or dependent variables).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information (e.g., age and gender) of the invention of Podoynitsina to further specify wherein the multiple pieces of demographic information further includes a residence region and a profession of the invention of Malmi et al. because doing so would allow the method to predict other demographic information such as age, location information, and occupation (see Malmi et al., Paragraph 0041). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Although Podoynitsina discloses normalizing the DFDS collected from the different platforms (e.g., preprocessing data obtained from the app store and webpages), Podoynitsina does not specifically disclose wherein the normalization of data includes rescaling rating related data to a uniform scale.
However, Zatorski et al. discloses for each group of the linked DFDS associated with each of the plurality of users, training, based on a portion of the group of the linked DFDS, embeddings representing trails of the user (Paragraph 0007, According to an embodiment of the disclosure, a method of recommending content to a user may include based on a content recommendation request received from a device of the user. A Method may include obtaining a feedback vector of the user for at least one of a plurality of pieces of content. A Method may include obtaining feedback information for the plurality of pieces of content used by a plurality of users. A Method may include generating user embedding vectors for the plurality of users and content embedding vectors for the plurality of pieces of content, based on the feedback information. A Method may include determining a plurality of user groups and central users of the plurality of user groups by grouping the plurality of users based on the user embedding vectors for the plurality of users. A Method may include inputting user embedding vectors for the central users and the content embedding vectors for the plurality of pieces of content to an artificial neural network model configured to estimate a preference degree for content. A Method may include obtaining preference vectors of the central users, which indicate estimated preference degrees of the central users for the plurality of pieces of content, the preference vectors being output by the artificial neural network model. A Method may include determining at least one piece of recommendation content to be recommended to the user, based on the feedback vector of the user and the preference vectors of the central users. A Method may include recommending the determined at least one piece of recommendation content to the user; Paragraph 0046, The user embedding vectors 1020 may include a user embedding vector 1021 of the first user 1001, a user embedding vector 1022 of the second user 1002, a user embedding vector 1023 of the third user 1003, a user embedding vector 1024 of a fourth user (not shown), a user embedding vector 1025 of a fifth user (not shown), and a user embedding vector 1026 of a sixth user (not shown), but are not limited thereto. The content embedding vectors 1030 may include a content embedding vector 1031 of first content, a content embedding vector 1032 of second content, a content embedding vector 1033 of third content, a content embedding vector 1034 of fourth content, and a content embedding vector 1035 of fifth content, but are not limited thereto; Paragraph 0047, Also, the server 1000 may train an artificial neural network model based on the user embedding vectors 1020 of the plurality of users and the content embedding vectors 1030 of the plurality of pieces of content in operation 1040. The artificial neural network model 1070 may be used to estimate a preference degree of a user for content; Paragraph 0188, The device 3200 may include a smartphone, a tablet personal computer (PC), a PC, a smart television (TV), a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, an electronic book terminal, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, or another mobile or non-mobile computing device, but is not limited thereto);
normalizing the linked DFDS collected from the different platforms, wherein ratings from the plurality of users are rescaled so that all rating related data are recorded using a uniform scale without changing relative evaluations from different users (Paragraph 0035, In the present specification, content may include multimedia content including movies, music, and the like, and content including various products purchasable from the Internet or an application downloadable from an application store; Paragraph 0074, According to an embodiment of the disclosure, a feedback vector of a user may be a vector indicating a preference degree of the user for at least one of a plurality of pieces of content. In detail, the preference degree of the user for at least one of the plurality of pieces of content may be determined based on a behavior or reaction of the user for the at least one of the plurality of pieces of content, and the preference degrees of the user for the plurality of pieces of content may be classified as “positive” or “negative”, or “positive”, “negative”, or “no preference”, but are not limited thereto. For example, the preference degrees of the user for the plurality of pieces of content may be classified as “positive” or “negative”, or “positive”, “negative”, or “no preference”, based on behaviors of the user clicking the plurality of pieces of content, scrolling the plurality of pieces of content, looking at the plurality of pieces of content for at least a certain period of time, scoring the plurality of pieces of content, downloading the plurality of pieces of content, streaming the plurality of pieces of content for at least a certain period of time, or watching previews of the plurality of pieces of content. Also, for example, a method of classifying the preference degrees of the user for the plurality of pieces of content as “positive” or “negative”, or “positive”, “negative”, or “no preference” may be stored in the server after being changed or pre-set by a processor of the server; Paragraph 0188, The device 3200 may include a smartphone, a tablet personal computer (PC), a PC, a smart television (TV), a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, an electronic book terminal, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, or another mobile or non-mobile computing device, but is not limited thereto; Examiner interprets “rating the plurality of pieces of content as positive or negative” as “normalizing the data to a uniform scale”);
generating an input vector for the user based on the group of the normalized DFDS including the trained embeddings, by using the trained embeddings to map the portion of the group of the linked DFDS to a sub-vector of the input vector, wherein the portion has a higher dimension and the sub-vector has a lower dimension (Paragraph 0096, The server may change the high-dimensional user embedding vectors for the plurality of users to the pre-set low-dimensional user embedding vectors for the plurality of users, through a principal component analysis (PCA)).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information of each of the plurality of users based on data received from multiple sources of the invention of Podoynitsina to further specify how the data received from multiple sources is embedded and normalized of the invention of Zatorski et al. because doing so would allow the method to input user embedding vectors for the central users and the content embedding vectors for the plurality of pieces of content to an artificial neural network model configured to estimate a preference degree for content (see Zatorski et al., Paragraph 0007). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Regarding claim 15 (Currently Amended), Podoynitsina discloses a system for prediction of demographics or interests, comprising (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age);
a perceptron neural network implemented by a processor and configured for training, based on data streams representing trails of a plurality of users as dynamic input via a perceptron neural network, a joint prediction model by dynamically learning based on the dynamic input, wherein the data streams are from different sources on different platforms providing data with respect to different types of identifications and different types of events; a joint model based demographic/interest prediction engine implemented by a processor and configured for receiving data from different sources (DFDS) having identifications included therein, wherein the DFDS relates to a plurality of users identifiable via the identifications (Page 45, Demographic Model, Once the information about topics is aggregated, we need to build a demographic model. Demographic model consist of several demographic classifies. In our case: age classifier, gender classifier and marital status classifier. In this work, we chose to use the deep learning approach using the Veles framework. Each classifier was built with neural network and optimized with genetic algorithm. The architecture of neural network is based on the multi-layer perceptron (Collobert and Bengio, 2004[18]); Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android plat form, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data. Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms);
linking DFDS associated with each of the plurality of users based on the identifications so that each group of the linked DFDS is under a same identification (Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android plat form, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data. Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms; Examiner notes that DFDS (e.g., web page data and application data) is associated to the same user (e.g., mobile device));
for each group of the linked DFDS associated with each of the plurality of users, training, based on a portion of the group of the linked DFDS, embeddings representing trails of the user (Page 44, Introduction, First we should build and train topic model with classifier of text data to specified categories (interests). The list of wanted categories can be given by content provider. The topic model categorizes the text extracted from Web pages. We build topic model with the Additive Regularization of Topic Models (ARTM) algorithm. Then we extract user interests using trained topic model. ARTM can be used not only for clustering, but for classification for a given list of categories. ARTM is based on generalization of two powerful algorithms: probabilistic latent semantic analysis (PLSA) [9] and latent Dirichlet allocation (LDA); Page 45, Document Aggregation, The resulting topic vector (or user interest vector) is used as feature vector for demographic model; Page 46, Demographic Model, We used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model; Page 46, Through the early tests, we chose the NN approach to build demographic prediction classifier. Currently different deep learning frameworks are available for training NNs: Caffe [13], Torch [14], Theano [15] and others. However, we used our custom deep learning framework (Veles) [16] because it was designed as a very flexible tool in terms of workflow construction, data extraction and preprocessing, visualization, with an additional advantage in the ease of porting of the resulting classifier to mobile devices; As stated in Paragraph 0038 of Applicant’s specification, embeddings may take data in a certain dimension as input and then generate an output vector with attributes characterizing the input in a different dimension. Therefore, based on broadest reasonable interpretation in light of the specification, Podoynitsina discloses “embeddings representing trails of the user” since it can reduce the size of an input feature vector. In this case, the input feature vector is the text extracted from webpages);
normalizing the linked DFDS collected from the different platforms, …; generating an input vector for the user based on the group of the normalized DFDS including the trained embeddings, by using the trained embeddings to map the portion of the group of the linked DFDS to a sub-vector of the input vector, wherein the portion has a higher dimension and the sub-vector has a lower dimension (Page 44, Introduction, First we should build and train topic model with classifier of text data to specified categories (interests). The list of wanted categories can be given by content provider. The topic model categorizes the text extracted from Web pages. We build topic model with the Additive Regularization of Topic Models (ARTM) algorithm. Then we extract user interests using trained topic model; Page 45, Preprocessing, The major source of our observation data is webpages browsed by the user. We preprocess the web pages as follows: remove HTML tags, perform stemming or lemmatization of every word, remove stop words, lowercase all characters and translate webpage content into a target languages; Page 45, Document Aggregation, The resulting topic vector (or user interest vector) is used as feature vector for demographic model; Page 46, Demographic Model, We used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model; Examiner interprets “preprocessing the data” as “normalizing the data”);
and generating an output vector by predicting, using the trained joint prediction model based on the input vector associated with the user, multiple pieces of demographic or interest information of the user (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age), wherein the input vector has a higher dimensionality than the output vector (Page 46, Demographic Model, We use several hyper-parameters of the neural network architecture: size of minibatch, number of layers, number of neurons in each layer, activation function, dropout, learning rate, weights de cay, gradient moment, standard deviation of weights, gradient de scent step, regularization coefficients, initial ranges of weights, number of examples per iteration. Genetic algorithm enables us to adjust these hyper-parameters. Also, we used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model), and wherein attributes included in the output vector are the predicted multiple pieces of demographic or interest information, wherein the predicted multiple pieces comprise a gender, an age, … (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age);
and distributing content to one or more target users selected based on the multiple pieces of demographic or interest information associated with each of the one or more target users (Page 44, Introduction, Our task is to build a demographical model, which will recognize demographic characteristics of user, such as gender, marital status and age. Such demographic information play a crucial role in personalized services and targeted advertising).
Although Podoynitsina discloses predicting multiple pieces of demographic or interest information (e.g., gender and age), Podoynitsina does not specifically disclose wherein the multiple pieces of demographic information further includes a residence region and a profession.
However, Malmi et al. discloses a system for prediction of demographics or interests comprising (Paragraph 0012, In one aspect, there is provided a method for predicting user demographics, and therewith e.g. user preferences, interests, purchase intent and/or other behavior or characteristics in view of providing personalized digital content, to be performed by at least one electronic apparatus, optionally a number of functionally connected servers potentially at least some which being operable in a cloud computing environment; Paragraph 0084, The terminal devices 104a, 104b, 104c, 104d, 104e, 104f and/or external devices/systems 114, 115, 116 directly or indirectly connected to the arrangement 114 for providing data thereto or obtaining data such as various deliverables therefrom, may generally contain similar hardware elements such as a processor, a memory and a communication interface): …
and wherein attributes included in the output vector are the predicted multiple pieces of demographic or interest information, wherein the predicted multiple pieces comprise a gender, an age, a residence region, and a profession (Paragraph 0041, The arrangement may be configured to contain or implement a number of functional modules to execute different method items, such as a data collector module for obtaining deterministic user data for modeling, a modeler for actually creating the model(s), a metering module for capturing (deterministic) usage statistics of apps for prediction purposes, and a predictor to estimate (predict) demographics based on the metered app usage data; Paragraph 0086, Model data collector 312 may be configured to receive (optionally by interrogation/pulling/survey mechanism) and manage (store, filter, combine, process, distribute) user data for modeling purposes. The data may contain hard, deterministic data having regard to a number of demographic characteristics (e.g. age, sex, marital status, race, income, language, country or other location information, occupation, and/or religion), application usage statistics and optional further, potentially behavioral data, which may be then utilized in creating a number of models associating e.g. app usage information (explanatory variables) with related predicted demographic characteristics (dependent variables) as explained herein. Further, a number of other characteristics such as behavioral characteristics may be included in the model as explanatory and/or dependent variables).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information (e.g., age and gender) of the invention of Podoynitsina to further specify wherein the multiple pieces of demographic information further includes a residence region and a profession of the invention of Malmi et al. because doing so would allow the method to predict other demographic information such as age, location information, and occupation (see Malmi et al., Paragraph 0041). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Although Podoynitsina discloses normalizing the DFDS collected from the different platforms (e.g., preprocessing data obtained from the app store and webpages), Podoynitsina does not specifically disclose wherein the normalization of data includes rescaling rating related data to a uniform scale.
However, Zatorski et al. discloses for each group of the linked DFDS associated with each of the plurality of users, training, based on a portion of the group of the linked DFDS, embeddings representing trails of the user (Paragraph 0007, According to an embodiment of the disclosure, a method of recommending content to a user may include based on a content recommendation request received from a device of the user. A Method may include obtaining a feedback vector of the user for at least one of a plurality of pieces of content. A Method may include obtaining feedback information for the plurality of pieces of content used by a plurality of users. A Method may include generating user embedding vectors for the plurality of users and content embedding vectors for the plurality of pieces of content, based on the feedback information. A Method may include determining a plurality of user groups and central users of the plurality of user groups by grouping the plurality of users based on the user embedding vectors for the plurality of users. A Method may include inputting user embedding vectors for the central users and the content embedding vectors for the plurality of pieces of content to an artificial neural network model configured to estimate a preference degree for content. A Method may include obtaining preference vectors of the central users, which indicate estimated preference degrees of the central users for the plurality of pieces of content, the preference vectors being output by the artificial neural network model. A Method may include determining at least one piece of recommendation content to be recommended to the user, based on the feedback vector of the user and the preference vectors of the central users. A Method may include recommending the determined at least one piece of recommendation content to the user; Paragraph 0046, The user embedding vectors 1020 may include a user embedding vector 1021 of the first user 1001, a user embedding vector 1022 of the second user 1002, a user embedding vector 1023 of the third user 1003, a user embedding vector 1024 of a fourth user (not shown), a user embedding vector 1025 of a fifth user (not shown), and a user embedding vector 1026 of a sixth user (not shown), but are not limited thereto. The content embedding vectors 1030 may include a content embedding vector 1031 of first content, a content embedding vector 1032 of second content, a content embedding vector 1033 of third content, a content embedding vector 1034 of fourth content, and a content embedding vector 1035 of fifth content, but are not limited thereto; Paragraph 0047, Also, the server 1000 may train an artificial neural network model based on the user embedding vectors 1020 of the plurality of users and the content embedding vectors 1030 of the plurality of pieces of content in operation 1040. The artificial neural network model 1070 may be used to estimate a preference degree of a user for content; Paragraph 0188, The device 3200 may include a smartphone, a tablet personal computer (PC), a PC, a smart television (TV), a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, an electronic book terminal, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, or another mobile or non-mobile computing device, but is not limited thereto);
normalizing the linked DFDS collected from the different platforms, wherein ratings from the plurality of users are rescaled so that all rating related data are recorded using a uniform scale without changing relative evaluations from different users (Paragraph 0035, In the present specification, content may include multimedia content including movies, music, and the like, and content including various products purchasable from the Internet or an application downloadable from an application store; Paragraph 0074, According to an embodiment of the disclosure, a feedback vector of a user may be a vector indicating a preference degree of the user for at least one of a plurality of pieces of content. In detail, the preference degree of the user for at least one of the plurality of pieces of content may be determined based on a behavior or reaction of the user for the at least one of the plurality of pieces of content, and the preference degrees of the user for the plurality of pieces of content may be classified as “positive” or “negative”, or “positive”, “negative”, or “no preference”, but are not limited thereto. For example, the preference degrees of the user for the plurality of pieces of content may be classified as “positive” or “negative”, or “positive”, “negative”, or “no preference”, based on behaviors of the user clicking the plurality of pieces of content, scrolling the plurality of pieces of content, looking at the plurality of pieces of content for at least a certain period of time, scoring the plurality of pieces of content, downloading the plurality of pieces of content, streaming the plurality of pieces of content for at least a certain period of time, or watching previews of the plurality of pieces of content. Also, for example, a method of classifying the preference degrees of the user for the plurality of pieces of content as “positive” or “negative”, or “positive”, “negative”, or “no preference” may be stored in the server after being changed or pre-set by a processor of the server; Paragraph 0188, The device 3200 may include a smartphone, a tablet personal computer (PC), a PC, a smart television (TV), a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, an electronic book terminal, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, or another mobile or non-mobile computing device, but is not limited thereto; Examiner interprets “rating the plurality of pieces of content as positive or negative” as “normalizing the data to a uniform scale”);
generating an input vector for the user based on the group of the normalized DFDS including the trained embeddings, by using the trained embeddings to map the portion of the group of the linked DFDS to a sub-vector of the input vector, wherein the portion has a higher dimension and the sub-vector has a lower dimension (Paragraph 0096, The server may change the high-dimensional user embedding vectors for the plurality of users to the pre-set low-dimensional user embedding vectors for the plurality of users, through a principal component analysis (PCA)).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information of each of the plurality of users based on data received from multiple sources of the invention of Podoynitsina to further specify how the data received from multiple sources is embedded and normalized of the invention of Zatorski et al. because doing so would allow the method to input user embedding vectors for the central users and the content embedding vectors for the plurality of pieces of content to an artificial neural network model configured to estimate a preference degree for content (see Zatorski et al., Paragraph 0007). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Regarding claims 2, 9, and 16 (Previously Presented), which are dependent of claims 1, 8, and 15, the combination of Podoynitsina, Malmi et al., and Zatorski et al. discloses all the limitations in claims 1, 8, and 15. Podoynitsina further discloses wherein the plurality of sources include: a plurality of platforms including …, mobile devices… (Page 46, Data Collection, The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data).
Although Podoynitsina receiving data from different sources (e.g., application data and web data), Podoynitsina does not specifically disclose wherein the different sources include: a plurality of platforms including desktop computers, laptop computers, mobile devices, and personal devices; and a plurality of applications operating on the plurality of platforms, wherein the personal devices include televisions, refrigerators, and audio devices.
However, Zatorski et al. discloses wherein the different sources include: a plurality of platforms including desktop computers, laptop computers, mobile devices, and personal devices; and a plurality of applications operating on the plurality of platforms, wherein the personal devices include televisions, refrigerators, and audio devices (Paragraph 0188, The device 3200 may include a smartphone, a tablet personal computer (PC), a PC, a smart television (TV), a mobile phone, a personal digital assistant (PDA), a laptop computer, a media player, a micro server, a global positioning system (GPS) device, an electronic book terminal, a digital broadcasting terminal, a navigation device, a kiosk, an MP3 player, a digital camera, a home appliance, or another mobile or non-mobile computing device, but is not limited thereto).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information based on data from different sources (e.g., application data and web data) of the invention of Podoynitsina to further specify wherein the different sources include a plurality of platforms including desktop computers, laptop computers, mobile devices, and personal devices of the invention of Zatorski et al. because doing so would allow the method to input user embedding vectors for the central users and the content embedding vectors for the plurality of pieces of content to an artificial neural network model configured to estimate a preference degree for content (see Zatorski et al., Paragraph 0007). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Regarding claims 3, 10, and 17 (Currently Amended), which are dependent of claims 2, 9, and 16, the combination of and Podoynitsina, Malmi et al., and Zatorski et al. discloses all the limitations in claims 2, 9, and 16. Podoynitsina further discloses wherein the DFDS comprises data of different types collected from the different sources, wherein the different types include: the identifications used in connection with the plurality of platforms and the plurality of applications for identifying users; event data recording activities of the plurality of users conducted with respect to the plurality of applications operating on the plurality of platforms (Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android plat form, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data. Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms; Examiner notes that DFDS (e.g., web page data and application data) is associated to the same user (e.g., mobile device));
the embeddings (Page 44, Introduction, First we should build and train topic model with classifier of text data to specified categories (interests). The list of wanted categories can be given by content provider. The topic model categorizes the text extracted from Web pages. We build topic model with the Additive Regularization of Topic Models (ARTM) algorithm. Then we extract user interests using trained topic model. ARTM can be used not only for clustering, but for classification for a given list of categories. ARTM is based on generalization of two powerful algorithms: probabilistic latent semantic analysis (PLSA) [9] and latent Dirichlet allocation (LDA); Page 45, Document Aggregation, The resulting topic vector (or user interest vector) is used as feature vector for demographic model; Page 46, Demographic Model, We used the genetic algorithm to select optimal features in input feature vector and reduce size of input feature vector of demographic model; Page 46, Through the early tests, we chose the NN approach to build demographic prediction classifier. Currently different deep learning frameworks are available for training NNs: Caffe [13], Torch [14], Theano [15] and others. However, we used our custom deep learning framework (Veles) [16] because it was designed as a very flexible tool in terms of workflow construction, data extraction and preprocessing, visualization, with an additional advantage in the ease of porting of the resulting classifier to mobile devices; As stated in Paragraph 0038 of Applicant’s specification, embeddings may take data in a certain dimension as input and then generate an output vector with attributes characterizing the input in a different dimension. Therefore, based on broadest reasonable interpretation in light of the specification, Podoynitsina discloses “embeddings representing trails of the user” since it can reduce the size of an input feature vector. In this case, the input feature vector is the text extracted from webpages));
and application graph data characterizing relationships among applications based on dynamics of application installation information, application usage levels, application activities, and correlations with other application usage patterns associated with the plurality of user (Page 44, Introduction, We used different types of data for demography prediction like call log, sms log, application usage data and so on. The demographic model consists of several (in our case 3) demographic classifiers. Demographic classifiers have to predict the age of the user (one of labels "0-18", «19- 21», «22 - 29, "30+"), gender (male or female) or marital status (married/not married) based on given feature vector; Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age; Page 46, Data Collection, Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store).
Regarding claims 4, 11, and 18 (Original), which are dependent of claims 3, 10, and 15, the combination of Podoynitsina, Malmi et al., and Zatorski et al. discloses all the limitations in claims 3, 10, and 15. Podoynitsina further discloses wherein each of the identifications in the DFDS is one of a user identification, a browser identification, and a device user identification (Page 46, Data Collection, To build robust demographic prediction models, we collected an extensive dataset with various available types of features from mobile users. To accomplish this task, we developed an entire system: 1) Mobile application, which is implemented on Android platform, periodically captures and save user activities on the mobile device with user permission and sends it to a server 2) Server that monitors and controls data collection. The collected mobile user data features were largely divided into three main categories as follows: call+sensors data, application and web data; It can be noted that the claim language is written in alternative form. The limitation taught by Podoynitsina is based on “a device user identification"), wherein the device user identification includes an identification for advertisers and/or an application identification (Page 44, Introduction, Our task is to build a demographical model, which will recognize demographic characteristics of user, such as gender, marital status and age. Such demographic information play a crucial role in personalized services and targeted advertising; Page 46, Data Collection, Call+sensors data consists of SMS and call log, battery status, cell tower data, light sensor status, location information, magnetic field and Wi-Fi data (some depending on availability). Application data includes such information as package name, time of installation, price, market name, and category name. Application market-related information is obtained from Google Play store, Amazon App store and Samsung Galaxy Apps store. Web data is obtained from various browsers (i.e. Google Chrome or Samsung Browser) by using Android platform content provider functions to get history. The history of browsing is then used to get textual (Web page content) information for further analysis by natural language processing (NLP) algorithms).
Regarding claims 5, 12, and 19 (Previously Presented), which are dependent of claims 1, 12, and 15, the combination of Podoynitsina, Malmi et al., and Zatorski et al. discloses all the limitations in claims 1, 12, and 15. Podoynitsina further discloses wherein the joint prediction model is derived to simultaneously predict multiple pieces of demographic or interest information (Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age).
Regarding claims 6 and 13 (Previously Presented), which are dependent of claims 5 and 12, the combination of Podoynitsina, Malmi et al., and Zatorski et al. discloses all the limitations in claims 5 and 12. Podoynitsina further discloses the gender is predicted as one of two genders; the age is predicted as one of a plurality of age groups; … (Page 44, Introduction, The demographic model consists of several (in our case 3) demographic classifiers. Demographic classifiers have to predict the age of the user (one of labels "0-18", «19- 21», «22 - 29, "30+"), gender (male or female) or marital status (married/not married) based on given feature vector; Page 46, Data Collection, The information collected from mobile phones of users can be used to predict many types of demographic parameters; however this work mainly focused on gender, marital status and age)
Although Podoynitsina discloses predicting multiple pieces of demographic or interest information (e.g., gender and age), Podoynitsina does not specifically disclose wherein the multiple pieces of demographic information further includes a residence region and a profession.
However, Malmi et al. further discloses wherein the gender is predicted as one of two genders; the age is predicted as one of a plurality of age groups; the residence region is predicted as one of a plurality of residence regions; and the profession is predicted as one of a plurality of professional categories (Paragraph 0041, The arrangement may be configured to contain or implement a number of functional modules to execute different method items, such as a data collector module for obtaining deterministic user data for modeling, a modeler for actually creating the model(s), a metering module for capturing (deterministic) usage statistics of apps for prediction purposes, and a predictor to estimate (predict) demographics based on the metered app usage data; Paragraph 0086, Model data collector 312 may be configured to receive (optionally by interrogation/pulling/survey mechanism) and manage (store, filter, combine, process, distribute) user data for modeling purposes. The data may contain hard, deterministic data having regard to a number of demographic characteristics (e.g. age, sex, marital status, race, income, language, country or other location information, occupation, and/or religion), application usage statistics and optional further, potentially behavioral data, which may be then utilized in creating a number of models associating e.g. app usage information (explanatory variables) with related predicted demographic characteristics (dependent variables) as explained herein. Further, a number of other characteristics such as behavioral characteristics may be included in the model as explanatory and/or dependent variables; Paragraph 0091, To enable implementation of a binary type of a solution, the demographic and potential other dependent variables may be first binarized when necessary. For example, an age-related variable may be established to cover two classes relating to ages between 18 and 32 years and ages between 33 and 100 years accordingly. Likewise, race variable could consist of white and non-white classes. The classes may be balanced, which facilitates comparing the predictability of different demographic variables, for example; Paragraph 0101, Generally, correlation between certain apps and a number of demographic characteristics was found to be strong. For example, the use of certain sports related apps or game apps was found to correlate well with male gender within an inspected age group whereas the presence of the aforesaid period tracking applications and some web stores in the list of used apps strongly implied a female user instead).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information (e.g., age and gender) of the invention of Podoynitsina to further specify wherein the multiple pieces of demographic information further includes a residence region and a profession of the invention of Malmi et al. because doing so would allow the method to predict other demographic information such as age, location information, and occupation (see Malmi et al., Paragraph 0041). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Claims 7, 14, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Podoynitsina (Podoynitsina, L., Romanenko, A., Kryzhanovskiy, K. and Moiseenko, A., 2017. Demographic prediction based on mobile user data. Electronic Imaging, 29, pp.44-47), in view of Malmi et al. (US 2018/0189660 A1), in further view of Zatorski et al. (US 2023/0100788 A1) and Priyadarshan et al. (US 2012/0041969 A1).
Regarding claims 7 and 14 (Previously Presented), which are dependent of claims 1 and 8, the combination of Podoynitsina, Malmi et al., and Zatorski et al. discloses all the limitations in claims 1 and 8. Although Podoynitsina discloses wherein the predicted demographic data is used for distributing content (Page 44, Introduction, personalized services and targeted advertising), Podoynitsina does not specifically disclose how the system is determining an affinity of each of the plurality of users with respect to the one or more targeting criteria based on the multiple demographic or interest information of the user.
However, Priyadarshan et al. discloses wherein the step of distributing content comprises: obtaining one or more targeting criteria of the content based on a description associated with the content, wherein the one or more targeting criteria are indicative of demographics or interests of intended targets (Paragraph 0205, In some embodiments, the delivery system 106 can combine the one or more characteristics through the use of one or more Boolean operators, such as AND, OR, or NOT. The Boolean operator can be applied to the user characteristics and/or the user characteristic values. For example, a custom targeted segment could specify the gender characteristic AND the age characteristic. In this case, when the delivery system 106 uses the custom targeted segment for content delivery, the delivery system 106 will only deliver content associated with this segment to users that satisfy both the gender and age requirements);
determining an affinity of each of the plurality of users with respect to the one or more targeting criteria based on the multiple demographic or interest information of the user (Paragraph 0224, For example, for a period of time the user might purchase all of their music from the Jazz category. Based on this contextual characteristic, the user might be assigned to a segment that targets Jazz listeners. At some later period, the user might begin to purchase music from the Blues category. These purchases will alter the value for the recent purchase frequency characteristic. When this occurs, the system 106 can re-analyze the set of user characteristics against the one or more rules. Based on the new analysis, the system 106 can update the segments to which the user is assigned. In this example, the user may no longer be assigned to the segment that targets Jazz listeners. In some configurations, the delivery system 106 can periodically re-infer and re-assign the users to targeted segments; Paragraph 0226, First, the delivery system 106 analyzes user characteristic data related to an identified user for behavioral patterns (3202). Behavioral patterns that can be identified based on the user characteristic data include for example patterns that can indicate a user's present or long-term intent or interest. Some examples include identifying the users propensity to convert or click on an item of invitational content; identifying when a user is about to, or is traveling; identifying when a user is researching a product or content, etc. From the analysis of the user characteristic data the system can infer interests in various products or user intent (3204) and categorize the user into a behavioral segment representative of that interest or intent (3206));
selecting the one or more target users based on their respective affinities (Paragraph 0224, For example, for a period of time the user might purchase all of their music from the Jazz category. Based on this contextual characteristic, the user might be assigned to a segment that targets Jazz listeners. At some later period, the user might begin to purchase music from the Blues category. These purchases will alter the value for the recent purchase frequency characteristic. When this occurs, the system 106 can re-analyze the set of user characteristics against the one or more rules. Based on the new analysis, the system 106 can update the segments to which the user is assigned. In this example, the user may no longer be assigned to the segment that targets Jazz listeners. In some configurations, the delivery system 106 can periodically re-infer and re-assign the users to targeted segments; Paragraph 0226, First, the delivery system 106 analyzes user characteristic data related to an identified user for behavioral patterns (3202). Behavioral patterns that can be identified based on the user characteristic data include for example patterns that can indicate a user's present or long-term intent or interest. Some examples include identifying the users propensity to convert or click on an item of invitational content; identifying when a user is about to, or is traveling; identifying when a user is researching a product or content, etc. From the analysis of the user characteristic data the system can infer interests in various products or user intent (3204) and categorize the user into a behavioral segment representative of that interest or intent (3206));
and transmitting the content to the selected one or more target users (Paragraph 0259, After prioritizing the segments, the delivery system 106 can use the ordered list to aid in selecting invitational content to deliver to the user. To illustrate, suppose a user is assigned to segments a, b, c, and d and the associated target objective for these segments is to maximize the CTR. After analyzing the characteristics of the user, the prioritizing module 128 determines the user is most likely to click through on content related to segment c followed by b, d, and a. When the delivery system 106 receives a request for invitational content for the user, the delivery system 106 will first attempt to deliver content associated with segment c. If no content exists for segment c, the delivery system 106 will go down the prioritized list and pick content associated with the next best segment).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information based on data from different sources (e.g., application data and web data) of the invention of Podoynitsina to further obtain one or more targeting criteria of the content based on a description associated with the content of the invention of Priyadarshan et al. because doing so would allow the method to deliver content associated with a specific segment (see Priyadarshan, Paragraph 0259). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Regarding claim 20 (Previously Presented), which is dependent of claim 1, the combination of Podoynitsina, Malmi et al., and Zatorski et al. discloses all the limitations in claim 1. Although Podoynitsina discloses wherein the predicted demographic data is used for distributing content (Page 44, Introduction, personalized services and targeted advertising), Podoynitsina does not specifically disclose how the system is determining an affinity of each of the plurality of users with respect to the one or more targeting criteria based on the multiple demographic or interest information of the user.
However, Priyadarshan et al. further discloses wherein the targeting-based content distribution engine is configured for distributing content by: obtaining one or more targeting criteria of the content based on a description associated with the content, wherein the one or more targeting criteria are indicative of demographics or interests of intended targets (Paragraph 0205, In some embodiments, the delivery system 106 can combine the one or more characteristics through the use of one or more Boolean operators, such as AND, OR, or NOT. The Boolean operator can be applied to the user characteristics and/or the user characteristic values. For example, a custom targeted segment could specify the gender characteristic AND the age characteristic. In this case, when the delivery system 106 uses the custom targeted segment for content delivery, the delivery system 106 will only deliver content associated with this segment to users that satisfy both the gender and age requirements);
determining an affinity of each of the plurality of users with respect to the one or more targeting criteria based on the multiple demographic/interest information of the user (Paragraph 0224, For example, for a period of time the user might purchase all of their music from the Jazz category. Based on this contextual characteristic, the user might be assigned to a segment that targets Jazz listeners. At some later period, the user might begin to purchase music from the Blues category. These purchases will alter the value for the recent purchase frequency characteristic. When this occurs, the system 106 can re-analyze the set of user characteristics against the one or more rules. Based on the new analysis, the system 106 can update the segments to which the user is assigned. In this example, the user may no longer be assigned to the segment that targets Jazz listeners. In some configurations, the delivery system 106 can periodically re-infer and re-assign the users to targeted segments; Paragraph 0226, First, the delivery system 106 analyzes user characteristic data related to an identified user for behavioral patterns (3202). Behavioral patterns that can be identified based on the user characteristic data include for example patterns that can indicate a user's present or long-term intent or interest. Some examples include identifying the users propensity to convert or click on an item of invitational content; identifying when a user is about to, or is traveling; identifying when a user is researching a product or content, etc. From the analysis of the user characteristic data the system can infer interests in various products or user intent (3204) and categorize the user into a behavioral segment representative of that interest or intent (3206));
selecting the one or more target users based on their respective affinities (Paragraph 0224, For example, for a period of time the user might purchase all of their music from the Jazz category. Based on this contextual characteristic, the user might be assigned to a segment that targets Jazz listeners. At some later period, the user might begin to purchase music from the Blues category. These purchases will alter the value for the recent purchase frequency characteristic. When this occurs, the system 106 can re-analyze the set of user characteristics against the one or more rules. Based on the new analysis, the system 106 can update the segments to which the user is assigned. In this example, the user may no longer be assigned to the segment that targets Jazz listeners. In some configurations, the delivery system 106 can periodically re-infer and re-assign the users to targeted segments; Paragraph 0226, First, the delivery system 106 analyzes user characteristic data related to an identified user for behavioral patterns (3202). Behavioral patterns that can be identified based on the user characteristic data include for example patterns that can indicate a user's present or long-term intent or interest. Some examples include identifying the users propensity to convert or click on an item of invitational content; identifying when a user is about to, or is traveling; identifying when a user is researching a product or content, etc. From the analysis of the user characteristic data the system can infer interests in various products or user intent (3204) and categorize the user into a behavioral segment representative of that interest or intent (3206));
and transmitting the content to the selected one or more target users (Paragraph 0259, After prioritizing the segments, the delivery system 106 can use the ordered list to aid in selecting invitational content to deliver to the user. To illustrate, suppose a user is assigned to segments a, b, c, and d and the associated target objective for these segments is to maximize the CTR. After analyzing the characteristics of the user, the prioritizing module 128 determines the user is most likely to click through on content related to segment c followed by b, d, and a. When the delivery system 106 receives a request for invitational content for the user, the delivery system 106 will first attempt to deliver content associated with segment c. If no content exists for segment c, the delivery system 106 will go down the prioritized list and pick content associated with the next best segment).
It would have been obvious to one ordinary skill in the art before the effective filing date to modify the method for predicting, via a perceptron neural network, multiple pieces of demographic or interest information based on data from different sources (e.g., application data and web data) of the invention of Podoynitsina to further obtain one or more targeting criteria of the content based on a description associated with the content of the invention of Priyadarshan et al. because doing so would allow the method to deliver content associated with a specific segment (see Priyadarshan, Paragraph 0259). Further, the claimed invention is merely a combination of old elements, and in combination each element would have performed the same function as it did separately, and one of ordinary skill in the art would have recognized that the results of the combination were predictable.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure.
Wang (Wang, S., Li, X., Kou, X., Zhang, J., Zheng, S., Wang, J. and Gong, J., 2021. Sequential recommendation through graph neural networks and transformer encoder with degree encoding. Algorithms, 14(9), p.263) – discloses a wide deep model. On the one hand, it extracted the features of input data through manual feature engineering, and at the same time, it used a deep neural network to extract complex high-order features and combined the two features for prediction. DeepFM combines factorization and deep learning techniques, using factorization to extract low-order crossover features of input data and deep neural networks to extract the high-order crossover features of input data. At the same time, it combines low-level and high-level features into the multilayer perceptron to predict the probability of the user interacting with the item (see at least Pages 3-4, 2.1.2 Recommendation Algorithm Based on Deep Learning).
Kang (US 2015/0120759 A1) – discloses user input related to preference of items or item sets may be used as input for supervised machine learning to train a user preference prediction algorithm, module, or the like. Supervised machine learning can include any suitable type, including neural networks, single-layered perceptron, multi-layer perceptron, decision trees, Radial Basis Function (RBF) networks, statistical learning algorithms (e.g., a Bayesian network), Support Vector Machine (SVM), Relevance Vector Machine (RVM), linear or logistic regression, and the like (see at least Paragraph 0052).
Shah et al. (US 2023/0052274 A1) – discloses a particular user that frequently searches for and/or views cat videos may be associated with a feature embedding representative of the class corresponding to cats. Thus, feature embeddings translate relatively high dimensional vectors of information (e.g., text strings, images, videos, etc.) into a lower dimensional space to enable the classification of different but similar objects (see at least Paragraph 0074).
Zhong (Zhong, Erheng, Ben Tan, Kaixiang Mo, and Qiang Yang. "User demographics prediction based on mobile data." Pervasive and mobile computing 9, no. 6 (2013): 823-837) – discloses to predict users’ demographics, including ‘‘gender’’, ‘‘job type’’, ‘‘marital status’’, ‘‘age’’ and ‘‘number of family members’’, based on mobile data, such as users’ usage logs, physical activities and environmental contexts (see at least Abstract).
Aggarwal (US 2024/0205479 A1) – discloses to initialize a predictive model configured to demographic predictions for content items, system server(s) 126 assigns to each node of a first plurality of nodes of the predictive model for predicting demographic information for content items, a respective weighted value that represents a respective demographic for each content item of a first plurality of content items. According to some aspects of this disclosure, the predictive model may include, but is not limited to, an iterative KNN model and/or the like configured, as described herein, to provide predictions for content items. According to some aspects of this disclosure, each content item of the first plurality of content items may be content items associated with known demographic data/information. For example, some of the content items of the first plurality of content items may be animated movies, tv shows, cartoons, and/or the like associated with an age-based demographic of children ages 5-13 years old, and some of the content items of the first plurality of content items may include mature content-related movies, tv shows, and/or the like associated with an age-based demographic of adults over 21 years old. According to some aspects of this disclosure, the respective demographics for the first plurality of content items include, but are not limited to, age-related demographics, gender-related demographics, location-based demographics, ethnicity-based demographics, sexual orientation-based demographics, and/or the like. According to some aspects of this disclosure, each content item of the first plurality of content items may be content items associated with any known demographic data/information (see at least Paragraph 0052).
Furlan et al. (US 11,023,953 B1) – discloses the sentiment analysis module 508 can convert that review score 518 to the normalized review score 520. The review score 518 refers to a value, text, or image indicating a sentiment, reaction, or a combination thereof. The review score 518 can be in the form of a number, character, image, emoji, or a combination thereof. If a review score 518 exists for a customer review 438, the sentiment analysis module 508 can generate the normalized review score 520 by performing a conversion of the review score 418 to the normalized review score 520. For example, if the review score 518 is an image or emoji indicating a sentiment based on images of “stars” wherein 1 star indicates a low/unfavorable sentiment or rating while 5 stars indicates a high/favorable sentiment or rating for the product 413, features 436, the further product, or a combination thereof, the sentiment analysis module 508 can convert or rescale the “star” ratings to the normalized review score 520, associated with the “star” rating. In another embodiment, if the review score 518 is based on a “thumbs up” or “thumbs down” rating system indicating how favorable or unfavorable the further customer 432 views the product 413, features 436, a further product, or a combination thereof, the sentiment analysis module 508 can convert or rescale the “thumbs up” and “thumbs down” rating to the normalized review score 520, associated with the “thumbs up” or “thumbs down” rating (see Column 20, lines 1-22).
Priyadarshan et al. (US 2012/0041969 A1) – discloses another way of deriving user characteristics is through a user's products. Products should be broadly thought of as any product or content that a user owns, has previously purchased, is considering purchasing, has viewed in an online store or advertisement, etc. From a collection of products the system can infer many user characteristics. For example, if the user owns an item of digital content targeted at teenagers, the system can infer that the user is a teenager; Paragraph 0074, For example, in the case of handheld communications devices, e.g. mobile phones, smart phones, tablets, or other types of user terminals connecting using multiple or non-persistent network sessions, multiple requests for content from such devices may be assigned to a same entry in the UUID database 116 (see at least Paragraph 0016).
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MARJORIE PUJOLS-CRUZ whose telephone number is (571)272-4668. The examiner can normally be reached Mon-Thru 7:30 AM - 5:00 PM.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Patricia H Munson can be reached at (571)270-5396. 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.
/M.P./Examiner, Art Unit 3624 /PATRICIA H MUNSON/Supervisory Patent Examiner, Art Unit 3624
Prosecution Insights