DETAILED ACTION
Claims 1-20 are presented for examination
This office action is in response to submission of application on 26-OCTOBER-2022.
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 Amendment
The amendment filed on 02-JANUARY-2026 in response to the non-final office action mailed 01-OCTOBER-2025 has been entered. Claims 1-20 remain pending in the application.
With regards to the 101 rejection, the rejection to claim 1 has not been overcome by the applicant’s amendments.
With regards to the 103 rejections, the applicant’s amendments to the claims have not overcome the rejections to claims 1-20 as the former prior art sufficiently teaches the newly amended claims.
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
an abstract idea (Abstract Idea) without significantly more.
Regarding claim 1, in Step 1 of the 101 analysis set forth in MPEP 2106, the claim recites a method for recommending content. A method is one of the four statutory categories of invention.
In Step 2a Prong 1 of the 101 analysis set forth in the MPEP 2106, the examiner has determined
that the following limitations recite a process that, under the broadest reasonable interpretation, covers
a mental process but for recitation of generic computer components:
Determining, by at least one processor of a server for recommending content to the user, 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; (one can mentally determine a plurality of groups based on the data assigned to it as a process of simply evaluating the data and making a determination based on that data)
Determining, by at least one processor of a server for recommending content to the user, 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 user; (one can mentally determine a piece of content to use based on given data as a process of simply evaluating the data and making a determination based on that data)
and recommending, by at least one processor of a server for recommending content to the user, the determined at least one piece of recommendation content to the user. (one can mentally make a recommendation/selection on a piece of content as a process of simply evaluating the content and making a determination based on that data)
determining similarity values between the estimated preference degrees of the central users for the plurality of pieces of content and a preference degree of the user for the plurality of pieces of content, based on the feedback vector of the user and the preference vectors of the central users; (one can mentally determine a similarity value based on given data as a process of simply evaluating the data and making a determination based on that data)
and adjusting a number of the plurality of user groups based on the determined similarity values (one can mentally adjust a value based on given data as a process of simply evaluating the data and making a determination based on that data)
determining the plurality of user groups based on a pre-set number of the plurality of user groups and Euclidean distances between the user embedding vectors for the plurality of users; (one can mentally determine a plurality of groups based on the data assigned to it as a process of simply evaluating the data and making a determination based on that data)
determining the central users of the plurality of user groups based on Euclidean distances between user embedding vectors for at least one user belonging to each of the determined plurality of user groups. (one can mentally determine a user from a group based on the data assigned to it as a process of simply evaluating the data and making a determination based on that data)
If claim limitations, under their broadest reasonable interpretation, covers performance of the limitations as a mental process but for the recitation of generic computer components, then it falls within the mental process grouping of abstract ideas. According, the claim “recites” an abstract idea.
In Step 2a Prong 2 of the 101 analysis set forth in MPEP 2106, the examiner has determined that the following additional elements do not integrate this judicial exception into a practical application:
A method of recommending content to a user, the method comprising: based on a content recommendation request received from a device of the user, (Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h))
obtaining a feedback vector of the user for at least one of a plurality of pieces of content; obtaining feedback information for the plurality of pieces of content used by a plurality of users; (Adding insignificant extra-solution activity (mere data gathering) to the judicial exception (MPEP 2106.05(g))
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; (In step 2A prong 2 generating a vector is a mere application of a computer tool (M.L. Model), which is not indicative of integration into a practical application. In step 2B, merely applying a computer tool is not indicative of significantly more.)
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; (Merely reciting the words “apply it” (or an equivalent) with the judicial exception (MPEP 2105.04(d)))
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; (Adding insignificant extra-solution activity (mere data gathering) to the judicial exception (MPEP 2106.05(g))
wherein the determining of the plurality of user groups and the central users of the plurality of user groups comprises: (Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h))
Since the claim does not contain any other additional elements that are indicative of integration into a practical application, the claim is “directed” to an abstract idea.
In Step 2b of the 101 analysis set forth in the 2019 PEG, the examiner has determined that the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
As discussed above, the additional element (viii) and (xiii) recites generally linking the use of the judicial exception to a particular technological environment or field of use, (x) and (xii) recites mere data gathering, (vi) recites a mere application of a computer tool, (xi) merely recites the recitation of “apply it” which is not indicative of significantly more. Considering the additional elements individually and in combination, and the claim as a whole, the additional elements do not provide significantly more than the abstract idea. Therefore, the claim is not patent eligible
Regarding claim 2, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 2 recites The method of claim 1, wherein the feedback vector of the user is a vector indicating a preference degree of the user for at least one of the plurality of pieces of content, and wherein the feedback information comprises a plurality of feedback vectors indicating preference degrees of the plurality of users for the plurality of pieces of content. (Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 3, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 3 recites The method of claim 1, wherein the generating of the user embedding vectors for the plurality of users and the content embedding vectors for the plurality of pieces of content comprises: generating user embedding vectors for the plurality of users and content embedding vectors for the plurality of pieces of content, which include the feedback information and errors equal to or less than a pre-set first threshold value, according to a pre-set standard. (In step 2A prong 2 generating a vector is a mere application of a computer tool (M.L. Model), which is not indicative of integration into a practical application. In step 2B, merely applying a computer tool is not indicative of significantly more.) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 5, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 5 recites The method of claim 1, wherein the determining of the at least one piece of recommendation content to be recommended to the user comprises: selecting a central user among the central users, based on the feedback vector of the user and the preference vectors of the central users; (In step 2A, prong 1, this recites an abstract idea but for recitation of generic computer components which is not indicative of integration into a practical application.) extracting a preference vector of the selected central user from among the preference vectors of the central users; (In step 2A, prong 1, this recites an abstract idea but for recitation of generic computer components which is not indicative of integration into a practical application.) and determining the at least one piece of recommendation content based on the extracted preference vector. (In step 2A, prong 1, this recites an abstract idea but for recitation of generic computer components which is not indicative of integration into a practical application.) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 7, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 7 recites The method of claim 6, wherein the adjusting of the number of the plurality of user groups comprises: selecting a central user among the central users, based on the feedback vector of the user and the preference vectors of the central users; identifying a similarity value of the selected central user from among the determined similarity values; and based on the identified similarity value being less than or equal to a pre-set second threshold value, increasing the number of the plurality of user groups. (In step 2A, prong 1, this recites an abstract idea but for recitation of generic computer components which is not indicative of integration into a practical application.) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 8, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 8 recites The method of claim 6, wherein the adjusting of the number of the plurality of user groups comprises: selecting a central user among the central users, based on the feedback vector of the user and the preference vectors of the central users; identifying a similarity value of the selected central user from among the determined similarity values; and based on the identified similarity value being greater than or equal to than a pre-set third threshold value, decreasing the number of the plurality of user groups. (In step 2A, prong 1, this recites an abstract idea but for recitation of generic computer components which is not indicative of integration into a practical application.) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 9, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 9 recites The method of claim 1, further comprising adjusting a number of the plurality of user groups, by comparing Euclidean distances between user embedding vectors for at least one user belonging to each of the plurality of user groups with a pre-set fourth threshold value. (Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 10, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 10 recites The method of claim 1, wherein the artificial neural network model comprises at least one of a generalized matrix factorization (GMF) model, a multi-layer perception (MLP) model, or a neural matrix factorization (NeuMF) model. (Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 11, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 11 recites The method of claim 1, wherein the artificial neural network model is trained by using, as training data, the user embedding vectors for the plurality of users and the content embedding vectors for the plurality of pieces of content. (Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 12, it is dependent upon claim 6, and thereby incorporates the limitations of, and corresponding analysis applied to claim 6. Further, claim 12 recites The method of claim 6, further comprising: generating a user embedding vector for the user based on the feedback information and the feedback vector of the user, and regenerating the user embedding vectors for the plurality of users and the content embedding vectors for the plurality of pieces of content; (In step 2A prong 2 generating a vector is a mere application of a computer tool (M.L. Model), which is not indicative of integration into a practical application. In step 2B, merely applying a computer tool is not indicative of significantly more.) and regrouping the user and the plurality of users based on the adjusted number of the plurality of user groups. (In step 2A, prong 1, this recites an abstract idea but for recitation of generic computer components which is not indicative of integration into a practical application.) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claim 20, it is dependent upon claim 1, and thereby incorporates the limitations of, and corresponding analysis applied to claim 1. Further, claim 20 recites A non-transitory computer-readable recording medium having recorded thereon a program for executing the method of claim 1 on a computer. (Generally linking the use of the judicial exception to a particular technological environment or field of use (MPEP 2106.05(h)) Since the claim does not recite additional elements that either integrate the judicial exception into a practical application, nor provide significantly more than the judicial exception, the claim is not patent eligible.
Regarding claims 13-19, they comprise of limitations similar to those of claims 1, 4-5, 9-11 respectively and therefore are rejected for similar rationale.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-3, 5, 7-8, 11-13, 15, 19, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over LIU (U.S. Pub. No. US 20210326674 A1) in view of DING (U.S. Pub. No. US 20200288205 A1) in view of BARBEHOEN (U.S. Pub. No. US 20040258311 A1) in view of LEE (U.S. Pub. No. US 20080310556 A1)
Regarding claim 1, LIU substantially teaches the claim including:
A method of recommending content to a user, the method comprising: based on a content recommendation request received from a device of the user, obtaining a feedback vector of the user for at least one of a plurality of pieces of content; ([0171] To ensure that content such as news information can be quickly exposed in real time, a recommendation system shown in FIG. 9 further includes a user feedback monitoring module. After one piece of recommendation content is pushed to a plurality of target users, many interactive users (for example, users who have clicked on the recommendation content) are generated every second. The user feedback monitoring module acquires a user click log 712 in real time, and collects a user who has recently clicked on the recommendation content as a seed user. Therefore, a seed user list corresponding to each piece of candidate recommendation content is updated in real time.
[0172] For example, the user feedback monitoring module updates, in real time, a seed user list and a seed user vector that are corresponding to candidate recommendation content in inventory. (the seed user vector comprises of the user content and the feedback of the user)) obtaining feedback information for the plurality of pieces of content used by a plurality of users; ([0172] For example, the user feedback monitoring module updates, in real time, a seed user list and a seed user vector that are corresponding to candidate recommendation content in inventory. To prevent a problem that a content recommendation server 114 becomes overburdened when performing subsequent clustering calculation on seed user vectors because too many users click on the candidate recommendation content in a short time, a seed user group corresponding to each piece of candidate recommendation content retains only three million users that have recently generated an interaction behavior. Certainly, a quantity of retained users for each piece of candidate recommendation content may be three million, or may be another quantity. A specific parameter is set by a technician. The interaction behavior includes at least one of clicking, reading, liking, commenting, and forwarding.) generating user embedding vectors for the plurality of users ([0026] a candidate content determining module, configured to determine n groups of seed user vectors according to the target user vector, each of the n groups of seed user vectors corresponding to a respective piece of candidate recommendation content, and n being a positive integer;)) determining, by at least one processor of a server for the recommending content to the user, 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; ([0025] an acquiring module, configured to acquire a target user vector of a target user; [0026] a candidate content determining module, configured to determine n groups of seed user vectors according to the target user vector, each of the n groups of seed user vectors corresponding to a respective piece of candidate recommendation content, and n being a positive integer; [0108] In an example, the content recommendation server 144 acquires a user feature of a user who has historically clicked on each piece of candidate recommendation content, and a user who has clicked on the candidate recommendation content is considered as a seed user of the candidate recommendation content. There may be L seed users in a group, and a seed user vector of each seed user is calculated according to a user feature of the seed user. The L seed user vectors are used for representing an interest feature of the candidate recommendation content. L is a positive integer.))
While LIU does teach gathering user vectors and feedback, it does not explicitly teach:
… and content embedding vectors for the plurality of pieces of content based on the feedback information; 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; 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; determining, by the at least one processor of the server for the recommending content to the user, 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 user; and recommending, by the at least one processor of the server for the recommending content to the user, the determined at least one piece of recommendation content to the user.
However, in analogous art that similarly gather user recommendation content data, DING teaches:
and content embedding vectors for the plurality of pieces of content based on the feedback information; ([0005] where the user interest information is corresponding to the feature of each convolutional layer; determining a first feature matrix based on the convolution of convolution kernel and the feature, where the convolution kernel includes the user interest information; (A matrix is a set of vectors. User interest information would be the content and feedback to the content. Thus the matrix is made of content vectors)) 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; ( [0073] In some embodiments, determining the user interest information includes that: the user interest information is determined by vectorizing the identifier based on a first embedding layer, where the user interest information is a matrix with an m*m size, and m refers to the width of the convolution kernel used by the each convolutional layer. [0074] In some embodiments, determining the user attention information includes that: the user attention information is determined by vectorizing the identifier based on a second embedding layer.[0075] The user attention information is a matrix with an m*m size, and m refers to a width of the convolution kernel used by the each convolutional layer. The second weight matrix of the second embedding layer is different from the first weight matrix of the first embedding layer.[0127] It is assumed that a training target at a training stage is a preference score in the form of real number. That is, a model including the convolutional neural network and the collaborative filtering neural network finally outputs a predicted score scored by the user to the multimedia resources, so that the fusion module then may output this preference score through the fully connected layers, and the electronic device makes a recommendation according to the obtained preference score in the form of real number. (the data input for the prediction score are the matrixes as disclosed. The matrixes are sets of vectors relating to the user and their interests, i.e. content, and the user and their attention, I.e. user data. Further, one can replace these matrixes with the users vectors and content vectors of LIU))) obtaining preference vectors of the central users, ([0114] In some embodiments, after the output vectors z.sub.i,s,t and w.sub.i,s,t are obtained, the two output vectors are fused. That is, the outputs of the convolutional layers are obtained by fusing the output vector z.sub.i,s,t by using the user attention weight. The outputs represent a preference degree of the recommended user to the output features of the corresponding convolutional layers, so that the outputs are also called the user preference data for the convolutional layers herein. That is, the user preference data are actually the matrixes. (the preference data is a matrix, which is a set of vectors)) 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; ([0114] In some embodiments, after the output vectors z.sub.i,s,t and w.sub.i,s,t are obtained, the two output vectors are fused. That is, the outputs of the convolutional layers are obtained by fusing the output vector z.sub.i,s,t by using the user attention weight. The outputs represent a preference degree of the recommended user to the output features of the corresponding convolutional layers, so that the outputs are also called the user preference data for the convolutional layers herein. That is, the user preference data are actually the matrixes. (the preference data is a matrix, which is a set of vectors) [0127] It is assumed that a training target at a training stage is a preference score in the form of real number. That is, a model including the convolutional neural network and the collaborative filtering neural network finally outputs a predicted score scored by the user to the multimedia resources, so that the fusion module then may output this preference score through the fully connected layers, and the electronic device makes a recommendation according to the obtained preference score in the form of real number.) determining, by the at least one processor of the server for the recommending content to the user, 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 user; and recommending, by the at least one processor of the server for the recommending content to the user, the determined at least one piece of recommendation content to the user. ([0020] In some embodiments, the processor is further configured to read and execute the executable instructions to: acquire a target weight matrix, where the target weight matrix provides weights corresponding to the N generated user preference data; weight and fuse the N generated user preference data based on the target weight matrix; and recommend the multimedia resources to the recommended user based on the weighted and fused user preference data.)
It would have been obvious to a person skilled in the art before the effective filing date of the
invention to have combined with DING‘s preference data and, with LIU‘s user and content vectors, with a reasonable expectation of success, a method for recommending content based on preference, as in DING, based on the user and content vectors, as found in LIU. A person of ordinary skill would have been motivated to reduce information overload (DING [0003]).
LIU further teaches:
determining similarity values between the estimated preference degrees of the central users for the plurality of pieces of content and a preference degree of the user for the plurality of pieces of content, based on the feedback vector of the user and the preference vectors of the central users; ([0111] The content recommendation server 144 calculates the similarity between the target user vector and each group of seed user vectors by using the look-alike model. Because each group of seed user vectors corresponds to one piece of candidate recommendation content, a degree of interest of the target user in each piece of candidate recommendation content may be evaluated by using a similarity of a group of seed user vectors corresponding to the candidate recommendation content. For example, a higher similarity indicates that the target user is more interested in the candidate recommendation content.)
While LIU, as modified by DING, does teach determining the similarity values between users, content, and preference, it does not explicitly teach:
and adjusting a number of the plurality of user groups based on the determined similarity values.
However, in analogous art that similarly measures similarity, BARBEHOEN teaches:
and adjusting a number of the plurality of user groups based on the determined similarity values. ([0029] By suitably selecting the feature similarity thresholds, similar new features are added to the group, i.e., the group's number of members and thus the group's strength increase. For example, the distance of a new feature from the calculated mean of the previously accepted members of a group can be used as a similarity value. A lower and/or upper threshold for this distance would in this example be designated a threshold. Another threshold consisting of a minimum number of object descriptive features (each assigned to corresponding groups) can be used. Less similar features are excluded from the group. A larger group contains more information on the object, which is described more precisely by the group or by the scattering values. )
It would have been obvious to a person skilled in the art before the effective filing date of the
invention to have combined with BARBEHOEN‘s data adjustment based on similarity and, with LIU‘s, as modified by DING, similarity calculation, with a reasonable expectation of success, a method for adjusting data based on similarity, as in BARBEHOEN, where the similarity is calculated using user and content data, as found in LIU, as modified by DING. A person of ordinary skill would have been motivated to improve model quality (BARBEHOEN [0007]).
While LIU, as modified by DING and BARBEHOEN, does teach finding user groups and the central user groups, it does not explicitly teach:
wherein the determining of the plurality of user groups and the central users of the plurality of user groups comprises: determining the plurality of user groups based on a pre-set number of the plurality of user groups and Euclidean distances between the user embedding vectors for the plurality of users; and determining the central users of the plurality of user groups based on Euclidean distances between user embedding vectors for at least one user belonging to each of the determined plurality of user groups.
However, in analogous art that similarly teaches grouping data, LEE teaches:
The method of claim 1, wherein the determining of the plurality of user groups and the central users of the plurality of user groups comprises: determining the plurality of user groups based on a pre-set number of the plurality of user groups and Euclidean distances between the user embedding vectors for the plurality of users; ([0053] The receiver estimates the transmit signal vectors based on all the transmittable x.sub.1, i.e., based on x.sub.1,1 through x.sub.1,16 as shown in FIG. 1A. More specifically, provided that x.sub.1 is transmittable 16 symbols, the receiver estimates 16 transmit signal vectors. x.sub.2, x.sub.3, and x.sub.4 are estimated using a QR Decomposition-Order Successive Interference Cancellation (QRD-OSIC) scheme, which will be described in further detail. Next, the receiver determines x.sub.1 candidate groups, i.e., determining candidate groups of the first phase by selecting 3 transmit signal vectors from the 16 transmit signal vectors. The receiver calculates the square Euclidean distance between the 16 transmit signal vectors and the receive signal vector, and selects 3 transmit signal vectors having the smallest square Euclidean distance value. The receiver calculates a per bit Log Likelihood Ratio (LLR) of x.sub.1 using the 16 square Euclidean distances acquired from the 16 transmit signal vectors.) and determining the central users of the plurality of user groups based on Euclidean distances between user embedding vectors for at least one user belonging to each of the determined plurality of user groups. ([0054] In FIG. 1B, the receiver estimates the other transmit signals by substituting all the transmittable x.sub.2, i.e., substituting x.sub.2,1 through x.sub.2,16 into the 3 selected x.sub.1 candidate groups among the 16 transmit signal vectors of FIG. 1A. The number of the estimated transmit signal vectors is 16 per transmittable x.sub.2 in each of the 3 x.sub.1 candidates, i.e., 48. The receiver calculates and updates per bit LLRs of x.sub.1 and x.sub.2 using 48 square Euclidean distances obtained from the 48 transmit signal vectors and the 16 square Euclidean distances of FIG. 1A. Hence, the per bit LLR of x.sub.1 is updated and the per bit LLR of x.sub.2 is newly acquired. With the 16 transmit signal vectors estimated in FIG. 1A and the 48 transmit signal vectors estimated in FIG. 1B, the square Euclidean distances of the 61 transmit signal vectors, excluding the overlapping transmit signal vectors, are used to calculate the LLR. The receiver determines candidate groups of x.sub.1-x.sub.2, i.e., determines candidate groups of the second phase by selecting 3 transmit signal vectors having the smallest square Euclidean distances amongst the 48 transmit signal vectors.)
It would have been obvious to a person skilled in the art before the effective filing date of the
invention to have combined with LEE‘s grouping of data and, with LIU‘s, as modified by DING, user and content vectors, with a reasonable expectation of success, a method for grouping data based on distance, as in LEE, where the data grouped are user vectors, as found in LIU, as modified by DING. A person of ordinary skill would have been motivated to lower complexity (LEE [0011]).
Regarding claim 2, DING further teaches:
The method of claim 1, wherein the feedback vector of the user is a vector indicating a preference degree of the user for at least one of the plurality of pieces of content, and wherein the feedback information comprises a plurality of feedback vectors indicating preference degrees of the plurality of users for the plurality of pieces of content. ([0114] In some embodiments, after the output vectors z.sub.i,s,t and w.sub.i,s,t are obtained, the two output vectors are fused. That is, the outputs of the convolutional layers are obtained by fusing the output vector z.sub.i,s,t by using the user attention weight. The outputs represent a preference degree of the recommended user to the output features of the corresponding convolutional layers, so that the outputs are also called the user preference data for the convolutional layers herein. That is, the user preference data are actually the matrixes. (here is taught a plurality of vectors, i.e. a matrix, which are made up of vectors that relate to the user that include preference degree. It would be obvious to use these vectors as the feedback vectors of LIU to gain the benefit of both))
Regarding claim 3, LIU teaches:
The method of claim 1, wherein the generating of the user embedding vectors for the plurality of users and the content embedding vectors for the plurality of pieces of content comprises: generating user embedding vectors for the plurality of users (([0025] an acquiring module, configured to acquire a target user vector of a target user;
[0026] a candidate content determining module, configured to determine n groups of seed user vectors according to the target user vector, each of the n groups of seed user vectors corresponding to a respective piece of candidate recommendation content, and n being a positive integer;))
DING further teaches:
and content embedding vectors for the plurality of pieces of content ([0005] where the user interest information is corresponding to the feature of each convolutional layer; determining a first feature matrix based on the convolution of convolution kernel and the feature, where the convolution kernel includes the user interest information; ),
LIU further teaches:
which include the feedback information ([0172] For example, the user feedback monitoring module updates, in real time, a seed user list and a seed user vector that are corresponding to candidate recommendation content in inventory. (the seed user vector comprises of the user content and the feedback of the user))) and errors equal to or less than a pre-set first threshold value, according to a pre-set standard. ([0187] A process of training the look-alike model 1141 may be referred to as look-alike learning 708. A user vector that is generated by user representation learning 707 and that can represent a user interest is used as input content for look-alike learning 708, historical clicks of the user are used as samples to train the user vector, and a click is used as a positive sample and a non-click is used as a negative sample, to adjust a weight of each neural network layer in the look-alike model 1141 according to error back propagation. [0206] Inputting a group of training samples into the user vector extraction model for training is considered as a training process. After the training process ends, the offline training server determines whether the current training process meets a training termination condition. The training termination condition includes, but is not limited to, the following two cases:
[0207] 1. The prediction error is less than a specified threshold;)
Regarding claim 5, LIU further teaches:
The method of claim 1, wherein the determining of the at least one piece of recommendation content to be recommended to the user comprises: selecting a central user among the central users, based on the feedback vector of the user and the preference vectors of the central users; ([0108] In an example, the content recommendation server 144 acquires a user feature of a user who has historically clicked on each piece of candidate recommendation content, and a user who has clicked on the candidate recommendation content is considered as a seed user of the candidate recommendation content. There may be L seed users in a group, and a seed user vector of each seed user is calculated according to a user feature of the seed user. The L seed user vectors are used for representing an interest feature of the candidate recommendation content. L is a positive integer.)
DING further teaches:
extracting a preference vector of the selected central user from among the preference vectors of the central users; ([0114] In some embodiments, after the output vectors z.sub.i,s,t and w.sub.i,s,t are obtained, the two output vectors are fused. That is, the outputs of the convolutional layers are obtained by fusing the output vector z.sub.i,s,t by using the user attention weight. The outputs represent a preference degree of the recommended user to the output features of the corresponding convolutional layers, so that the outputs are also called the user preference data for the convolutional layers herein. That is, the user preference data are actually the matrixes.) and determining the at least one piece of recommendation content based on the extracted preference vector. ([0020] In some embodiments, the processor is further configured to read and execute the executable instructions to: acquire a target weight matrix, where the target weight matrix provides weights corresponding to the N generated user preference data; weight and fuse the N generated user preference data based on the target weight matrix; and recommend the multimedia resources to the recommended user based on the weighted and fused user preference data.)
Regarding claim 7, LIU further teaches:
The method of claim 6, wherein the adjusting of the number of the plurality of user groups comprises: selecting a central user among the central users, based on the feedback vector of the user and the preference vectors of the central users; identifying a similarity value of the selected central user from among the determined similarity values; ([0100] The content recommendation server 114 calculates a similarity between a seed user vector of each piece of candidate recommendation content and the target user vector, where the similarity is used for indicating a degree of interest of the target user in the candidate recommendation content. [0110] The look-alike model is used for calculating the similarity based on an attention mechanism. The look-alike model can select a seed user vector that has a more reference value for the target user from a group of seed user vectors based on the attention mechanism, to perform similarity calculation.)
BARBEHOEN further teaches:
and based on the identified similarity value being less than or equal to a pre-set second threshold value, increasing the number of the plurality of user groups. ([0029] By suitably selecting the feature similarity thresholds, similar new features are added to the group, i.e., the group's number of members and thus the group's strength increase. For example, the distance of a new feature from the calculated mean of the previously accepted members of a group can be used as a similarity value. A lower and/or upper threshold for this distance would in this example be designated a threshold. Another threshold consisting of a minimum number of object descriptive features (each assigned to corresponding groups) can be used. Less similar features are excluded from the group. A larger group contains more information on the object, which is described more precisely by the group or by the scattering values. )
Regarding claim 8, LIU further teaches:
The method of claim 6, wherein the adjusting of the number of the plurality of user groups comprises: selecting a central user among the central users, based on the feedback vector of the user and the preference vectors of the central users; identifying a similarity value of the selected central user from among the determined similarity values; ([0100] The content recommendation server 114 calculates a similarity between a seed user vector of each piece of candidate recommendation content and the target user vector, where the similarity is used for indicating a degree of interest of the target user in the candidate recommendation content. [0110] The look-alike model is used for calculating the similarity based on an attention mechanism. The look-alike model can select a seed user vector that has a more reference value for the target user from a group of seed user vectors based on the attention mechanism, to perform similarity calculation.)
BARBEHOEN further teaches:
and based on the identified similarity value being greater than or equal to than a pre-set third threshold value, decreasing the number of the plurality of user groups. ([0029] Another threshold consisting of a minimum number of object descriptive features (each assigned to corresponding groups) can be used. Less similar features are excluded from the group. A larger group contains more information on the object, which is described more precisely by the group or by the scattering values.)
Regarding claim 11, LIU further teaches:
The method of claim 1, wherein the artificial neural network model is trained by using, as training data, the user embedding vectors for the plurality of users and the content embedding vectors for the plurality of pieces of content. ([0190] Step 1001: Acquire first training samples, the first training samples including user features of a plurality of sample users and interaction records of the plurality of sample users for sample content.
[0191] The first training samples include the user features of the sample users, an embedding vector of the sample content, and the interaction records of the sample users for the sample content. The interaction record includes at least one of click, like, comment, and forward.)
Regarding claim 12, LIU further teaches:
The method of claim 6, further comprising: generating a user embedding vector for the user based on the feedback information and the feedback vector of the user, ([0191] The first training samples include the user features of the sample users, an embedding vector of the sample content, and the interaction records of the sample users for the sample content. The interaction record includes at least one of click, like, comment, and forward.
[0192] Step 1002: Input a user feature of a first sample user into the user vector extraction model to obtain a first user vector.) and regenerating the user embedding vectors for the plurality of users and the content embedding vectors for the plurality of pieces of content; and regrouping the user and the plurality of users based on the adjusted number of the plurality of user groups. ([0189] FIG. 10 shows a flowchart of a training method for a user vector extraction model according to an exemplary embodiment of this application. The method may be performed by a content recommendation server 114 or an additional offline training server. The user vector extraction model includes an embedding layer, a merging layer, and a full connection layer, where the merging layer is a neural network layer based on a self-attention mechanism. The training method includes the following steps:
[0190] Step 1001: Acquire first training samples, the first training samples including user features of a plurality of sample users and interaction records of the plurality of sample users for sample content.
[0191] The first training samples include the user features of the sample users, an embedding vector of the sample content, and the interaction records of the sample users for the sample content. The interaction record includes at least one of click, like, comment, and forward.
[0192] Step 1002: Input a user feature of a first sample user into the user vector extraction model to obtain a first user vector. (the steps of generating as shown in LIU is iterative as a result of being done during training. Thus, the generation and grouping will be redone every iteration))
Regarding claim 20, LIU further teaches:
A non-transitory computer-readable recording medium having recorded thereon a program for executing the method of claim 1 on a computer. ([0306] An embodiment of this application further provides a non-transitory computer-readable storage medium, storing at least one instruction, at least one program, a code set, or an instruction set, the at least one instruction, the at least one program, the code set, or the instruction set being loaded and executed by a processor, to implement the foregoing content recommendation method, the training method for a user vector extraction model, or the training method for a look-alike model.)
Regarding claim 13, it comprises of limitations similar to those of claim 1 and is therefore rejected for similar rationale. Regarding claim 15, it comprises of limitations similar to those of claim 5 and is therefore rejected for similar rationale. Regarding claim 19, it comprises of limitations similar to those of claim 11 and is therefore rejected for similar rationale.
Claims 9 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over LIU (U.S. Pub. No. US 20210326674 A1), DING (U.S. Pub. No. US 20200288205 A1), BARBEHOEN (U.S. Pub. No. US 20040258311 A1), LEE (U.S. Pub. No. US 20080310556 A1) in further view of GEFFEN (U.S. Pub. No. US 20210304103 A1)
While LIU, as modified by DING, does teach claim 1, which claim 9 is dependent upon, it does not explicitly teach:
The method of claim 1, further comprising adjusting a number of the plurality of user groups, by comparing Euclidean distances between user embedding vectors for at least one user belonging to each of the plurality of user groups with a pre-set fourth threshold value.
However, in analogous art that similar measures the distance between vectors, GEFFEN teaches:
The method of claim 1, further comprising adjusting a number of the plurality of user groups, by comparing Euclidean distances between user embedding vectors for at least one user belonging to each of the plurality of user groups with a pre-set fourth threshold value. ([0042] Compute the Euclidean distance (or other means of comparison) between the vector of the first agent's (Agent_ID) attributes and the vector of the given agent's attributes. If a distance is below a predefined threshold or otherwise within a predefined range, indicating an acceptable level of similarity between the first agent and the given agent, the processor may add this agent to the control group of agents. For example, a distance of 1.2 means the two vectors are more similar than a distance of 12.3.)
It would have been obvious to a person skilled in the art before the effective filing date of the
invention to have combined with GEFFEN‘s data adjustment based on similarity and, with LIU‘s, as modified by DING, user vectors, with a reasonable expectation of success, a method for adjusting data based on similarity, as in GEFFEN, where the data modified is user data, as found in LIU, as modified by DING. A person of ordinary skill would have been motivated to improve effectiveness (GEFFEN [0004]).
Regarding claim 17, it comprises of limitations similar to those of claim 9 and is therefore rejected for similar rationale.
Claims 10 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over LIU (U.S. Pub. No. US 20210326674 A1), DING (U.S. Pub. No. US 20200288205 A1), BARBEHOEN (U.S. Pub. No. US 20040258311 A1), LEE (U.S. Pub. No. US 20080310556 A1) in further view of ZARKOV (U.S. Pub. No. US 20200050988 A1)
While LIU, as modified by DING, does teach claim 1 which claim 10 is dependent upon, it does not explicitly teach:
The method of claim 1, wherein the artificial neural network model comprises at least one of a generalized matrix factorization (GMF) model, a multi-layer perception (MLP) model, or a neural matrix factorization (NeuMF) model.
However, in analogous art that similarly employs a neural network model, ZARKOV teaches:
The method of claim 1, wherein the artificial neural network model comprises at least one of a generalized matrix factorization (GMF) model, a multi-layer perception (MLP) model, or a neural matrix factorization (NeuMF) model. ([0020] Clause 10: A system for implementing a hybrid deep neural network model to determine a market strategy, comprising: at least one processor programmed or configured to: generate a first model comprising a generalized matrix factorization model)
It would have been obvious to a person skilled in the art before the effective filing date of the
invention to have combined with ZARKOV‘s model and, with LIU‘s, as modified by DING, method, with a reasonable expectation of success, a GMF model, as in ZARKOV, which executes a method for recommending content, as found in LIU, as modified by DING. A person of ordinary skill would have been motivated to improve effectiveness (ZARKOV [0005]).
Regarding claim 18, it comprises of limitations similar to those of claim 10 and is therefore rejected for similar rationale.
Response to Arguments
Applicant’s arguments filed 02-JANUARY-2026 have been fully considered, but they are found to be non-persuasive
With regards to the applicant’s remarks regarding the 101 rejection towards an abstract idea, the applicant argues that the amendments to claim 1 overcome the rejection of determining and selecting users to make a recommendation
The 2019 Revised Patent Subject Matter Eligibility Guidance indicates that "a claim is not 'directed to' a judicial exception, and thus is patent eligible, if the claim as a whole integrates the recited judicial exception into a practical application of that exception." In this case, independent claims 1 and 13, as a whole, integrate the recited recommendation concept into a practical application by providing recommendation performance, including adjusting the number of user groups in view of processing time, memory usage, and estimation error (difference value), thereby improving computational efficiency while maintaining or improving recommendation quality. Here, the present claims integrate the alleged judicial exception into a practical application by reciting a specific recommendation architecture that improves the functioning of the server and the overall recommendation system. In particular, the claims include (i) determining, by at least one processor of a server for the recommending content to the user, a plurality of user groups and central users of the plurality of user groups, (ii) adjusting a number of the plurality of user groups based on the determined similarity values, and (iii) determining the central users of the plurality of user groups based on Euclidean distances between user embedding vectors for at least one user belonging to each of the determined plurality of user groups.
With regards to this argument, the examiner acknowledges the addition of the adjustment of the user group numbers and its importance to the specification. However, it should be noted that the act of adjusting a number, in of itself, is an abstract idea. To be integrated into a practical application, the claim must recite more than an abstract idea and an abstract idea itself cannot integrate the claim into the practical application. According to MPEP 2106(a)(II), “To show that the involvement of a computer assists in improving the technology, the claims must recite the details regarding how a computer aids the method, the extent to which the computer aids the method, or the significance of a computer to the performance of the method. Merely adding generic computer components to perform the method is not sufficient. Thus, the claim must include more than mere instructions to perform the method on a generic component or machinery to qualify as an improvement to an existing technology. “ The improvement cannot simply lay in the abstract idea itself. Therefore, the examiner maintains that determining users and adjusting a number are abstract ideas.
With regards to the applicant’s remarks regarding the 103 rejection in the non-final action, the applicant argues that the prior art does not teach the newly amended claim 1. The examiner acknowledges this argument and has adjusted claim 1 to include the newly amended claim limitations from claims 4 and 6. However, applicant’s arguments regarding the newly added limitations to claim 1 have been found non-persuasive.
Applicant submit that DING does not teach or suggest claim 1. In particular, DING (see, e.g., paragraphs [0073]-[0075] and [0127]) refers to "user interest information" and "user attention information," but does not expressly teach or suggest grouping users and, determining a central user per group, and performing neural-network-based preference estimation using only the central users' embedding vectors as inputs.
With regards to this argument, it should be remembered that, in a 103 rejection, the limitations can be rejected by a combination of multiple prior arts. While DING does not explicitly teach the grouping and central users per group, it is combined with LIU which does teach those limitations. It would be obvious to one of ordinary skill in the art to use the methodology of DING in combination with the central users and user groups of LIU to achieve the claim limitations. LIU has already taught grouping users and determining the central user, DING has no burden to reteach those limitations.
The Office Action cites Lee (U.S. 2008/0310556 A1) for limitation of original claim 4 (added to the present claim 1) of determining a central user based on Euclidean distance. Applicants respectfully disagree. Lee is directed to wireless receiver signal detection (e.g., QRD- OSIC) and uses (square) Euclidean distance to select a small set of candidate transmit signal vectors and to compute/update per-bit LLRs by comparing hypothesized transmit vectors to a received signal vector. See Lee at [0053]-[0054]. This is fundamentally different from the present claims. In an exemplary embodiment, there is a central user as a central user within a user group in a recommendation system based on distances among user embedding vectors. Lee does not teach or suggest grouping user embedding vectors, determining a central user per group, or using such a central user for recommendation. Accordingly, Lee fails to disclose or render obvious original claim 4, and reliance on Lee does not provide a proper basis for the rejection.
With regards to this argument, LEE is not mapped to limitations claiming grouping user embedding vectors, determining a central user per group, or using a central user for recommendation. Rather, these limitations are required to be taught prior due to LEE inclusion in claim. These limitations have already been taught by LIU which is used in combination with LEE to achieve the limitations that LEE is mapped. As such, LEE in combination of LIU has already taught these limitations and is using them as basis for the limitation LEE is mapped to. LEE has no burden to reteach these limitations on its own.
The Office Action asserts that Barbehoen (U.S. 2004/0258311 A1) discloses the limitation of original claim 6 (added to the present claim 1) relating to adjusting the number of user groups. Applicants respectfully disagree. Barbehoen at most describes tuning a similarity threshold such that by suitably selecting the feature similarity thresholds, similar new features are added to the group, i.e., the group's number of members and thus the group's strength increase. In other words, Barbehoen concerns controlling the number of group membership/strength via thresholding, not adjusting the number of groups as required by claim 6. Accordingly, Barbehoen fails to disclose or suggest the original claim 6 limitation (added to the present claim 1).
With regards to this argument, BARBEHOEN does teach the limitations set forth. Specifically, BARBEHOEN teaches: “[0029] By suitably selecting the feature similarity thresholds, similar new features are added to the group, i.e., the group's number of members and thus the group's strength increase.” Specifically, as the threshold grows, new group members, i.e. features, are added which increase the strength of the group. This is by definition an adjustment. When in combination with LIU and DING, one could use the threshold similarity, i.e. similarity value, on the user groups as taught by LIU to achieve the claimed limitations. It should be noted that BARBEHOEN specifically states that the features are added as the threshold grows. So, if the features of BARBEHOEN are interpreted as the groups from LIU, the number of groups grow as the threshold/value increases.
Conclusion
THIS ACTION IS MADE FINAL. 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 SKIELER A KOWALIK whose telephone number is (571)272-1850. The examiner can normally be reached 8-5.
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, Mariela D Reyes can be reached at (571)270-1006. 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.
/SKIELER ALEXANDER KOWALIK/Examiner, Art Unit 2142 /Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
Prosecution Insights