Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
This action is in response to the amendments filed 02/10/2026. Claims 1, 8, 13, and 20 have been amended, claims 3 and 15 have been cancelled, claims 21-22 have been added. Claims 1-2, 4-14, and 16-22 are currently pending.
Response to Arguments
Claims 3 and 15 have been cancelled, therefore the rejections of claim 3 and 15 no longer stand.
Applicant’s arguments regarding the 101 rejection have been fully considered but they are not persuasive. Applicant argues that training the model “based on the sampled probability value” amounts to a “specific requirement” of the claims that reflects improvements in the operation of a computing system. Examiner respectfully disagrees and notes that the claim limitation reads “training the machine learning model to rank the plurality of entities for presentation via the online system based on the sampled probability value”. While one interpretation of this limitation could be that the training is based on the sampled probability value, the broadest reasonable interpretation of this limitation also includes wherein ranking the plurality of entities is based on the sampled probability value. A person could rank observed entities in their mind based on observed or mentally determined sampled probability values. Examiner also disagrees with Applicant’s characterization of an improvement provided by the claims, as at least paragraph [0013] Applicant’s specification states “These improvements to the entity selection and ranking processes improve the quality of data used by the machine learning model, and therefore improve the likelihood that high quality entities will be selected and highly ranked by the machine learning model in the future”..
Further, Examiner respectfully disagrees with Applicant’s comparison of the claims to Desjardins, as the claims in Desjardins were interpreted as being directed to an “improvement to how the machine learning model itself operates”. Examiner notes that the claims recite limitations directed to analyzing and curating data that is used by a machine learning model in a training process. While improving the content of training data may improve the output of a machine learning mode, it does not improve the structure or the operations performed by the machine learning model itself. Applicant has not claimed any particular machine learning model nor training steps that would indicate to one of ordinary skill the technical, computer implemented steps required to train this model to rank a plurality of entities. Applicant has not pointed to any claim limitations that would distinguish this ranking process from the way a person could mentally rank entities in their mind, as merely implemented by a generic computer component. Applicant’s claims are more suitably compared to Recentive Analytics, Inc v. Fox Corp., No. 23-2437 (Fed. Cir. 2025), as the claims merely recite use of a machine learning model in a new environment (see page 13 of Recentive). Page 15 of Recentive further states “the claimed methods are not rendered patent eligible by the fact that (using existing machine learning technology) they perform a task previously undertaken by a human with greater speed and efficiency than could previously be achieved. Whether the issue is raised at step one or step two, the increased speed and efficiency resulting from use of computers (with no improve computer technique) do not themselves create eligibility”. Similarly, Applicant has argued that the claims reflect an improvement to a data preprocessing or curation task that one of ordinary skill in the art would recognize was previously performed by data scientists, but Applicant has not shown how the claims reflect an improved machine learning model or other computer technique. The 101 rejections have been updated to include the amended limitations and to clarify the reasoning given for the limitations that were not amended where necessary.
Applicant’s arguments regarding the prior art rejection have been fully considered but are moot because of the new ground(s) of rejection. Applicant argues that the cited prior art does not teach “a variance determined using a number of followers of the entity”. Examiner notes that the Wang reference has been brought in to teach this limitation as amended, in combination with the Li and Kenthapadi references. The prior art rejections have been updated to include the amended limitations and to clarify the reasoning given for the limitations that were not amended.
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 1-2, 4-14, and 16-22 are rejected under 35 U.S.C. 101. Claims 1-2 and 4-7 are directed to a method, claims 8-12 are directed to a separate method, claims 13-14 and 16-19 are directed to a system, and claims 20-22 are directed to a separate system; therefore, claims 1-2, 4-14, and 16-22 fall within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter). However, claims 1-2, 4-14, and 16-22 fall within the judicial exception of an abstract idea, specifically the abstract ideas of “Mental Processes” (including observation, evaluation, and opinion) and “Mathematical Concepts (including mathematical calculations and relationships)”.
Claim 1:
Claim 1 is directed to a method; therefore, the claim does fall within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Claim 1 recites the following abstract ideas:
generating a reward score for an entity of a plurality of entities (mental step directed to observation, evaluation – a person could generate a reward score for an observed entity in their mind);
determining a rate distribution for the entity using the reward score (mental step directed to observation, evaluation – a person could determine a rate distribution for an observed entity using a mentally observed or determined reward score in their mind);
generating a sampled rate value for the entity by sampling the rate distribution (mental step directed to observation, evaluation – a person could sample a rate distribution in their mind to generate a sample rate value for an observed entity in their mind);
generating a probability score for a pair of the entity and a user using the sampled rate value (mental step directed to observation, evaluation – a person could use a mentally observed or determined sampled rate value to generate a probability score for an observed entity and associated user in their mind);
determining a probability distribution for the pair using the probability score, wherein the probability distribution has a mean that is the probability score and a variance determined using a number of followers of the entity (mental step directed to observation, evaluation – a person could determine a probability distribution in their mind with an observed or mentally determined mean as a probability score and an observed or mentally determined variance using a number of followers of an entity);
generating a sampled probability value for the pair by sampling the probability distribution (mental step directed to observation, evaluation – a person could sample a probability distribution to generate a sampled probability value for an observed entity and associated user in their mind);
wherein the sampled probability value is generated according to the mean and variance of the probability distribution (mental step directed to observation, evaluation – a person could generate a sampled probability value in their mind based on an observed or mentally determined mean and variance).
Claim 1 recites the following additional elements:
training the machine learning model to rank the plurality of entities for presentation via the online system based on the sampled probability value. As the claim does not recite any particular technical details regarding the machine learning model nor any training process, the machine learning model is interpreted as a generic computer component and training this model to rank entities using a sampled probability value is interpreted as merely using this generic computer component to apply a mental step associated with ranking entities using an observed or determined sampled probability value. Wherein the entities are ranked for presentation via the online system is interpreted as the intended use of the ranking, though Examiner notes that actively presenting ranked entities via an online system would be interpreted as transmitting data over a network.
These additional elements do not integrate the abstract idea into a practical application or amount to significantly more than the abstract idea (see MPEP 2106.05(d)(II) and MPEP 2106.05(f)).
Claim 8:
Claim 8 is directed to a method; therefore, the claim does fall within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Claim 8 recites the following abstract ideas:
generating a probability score for a pair of the entity and a user using the sampled rate value, wherein the probability score comprises at least one of a follow probability, a utility probability, or a create probability (mental step directed to observation, evaluation – a person could use a mentally observed or determined sampled rate value to generate a follow, utility, or create probability score for an observed entity and associated user in their mind);
determining a probability distribution for the pair using the probability score, wherein the probability distribution has a mean that is the probability score and a variance determined using a number of followers of the entity (mental step directed to observation, evaluation – a person could determine a probability distribution in their mind with an observed or mentally determined mean as a probability score and an observed or mentally determined variance using a number of followers of an entity);
sampling the probability distribution to generate a sampled probability value for the pair, wherein the sampled probability value is based on the mean and the variance of the probability distribution (mental step directed to observation, evaluation – a person could sample a probability distribution in their mind to generate a sampled probability value for an observed entity and associated user based on an observed or mentally determined mean and variance);
Claim 8 recites the following additional elements:
receiving a sampled rate value for an entity of a plurality of entities; and training the machine learning model to rank the plurality of entities for presentation via the online system using the sampled probability value. Receiving a sampled rate value for an entity is interpreted as well-understood, routine, conventional activity directed to receiving data over a network. As the claim does not recite any particular technical details regarding the machine learning model nor any training process, the machine learning model is interpreted as a generic computer component and training this model to rank entities using a sampled probability value is interpreted as merely using this generic computer component to apply a mental step associated with ranking entities using an observed or determined sampled probability value. Wherein the entities are ranked for presentation via the online system is interpreted as the intended use of the ranking, though Examiner notes that actively presenting ranked entities via an online system would be interpreted as transmitting data over a network.
These additional elements do not integrate the abstract idea into a practical application or amount to significantly more than the abstract idea (see MPEP 2106.05(d)(II) and MPEP 2106.05(f)).
Claim 13 is a system claim and its limitation is included in claim 1. The only difference is that claim 13 requires a system, which is interpreted as a generic computer component merely used to apply the abstract ideas as identified in the analysis of claim 1 (see MPEP 2106.05(f)). Therefore, claim 13 is rejected for the same reasons as claim 1.
Claim 20:
Claim 20 is directed to a system; therefore, the claim does fall within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Claim 20 recites the following abstract ideas:
determine whether to select the entity using the sampled rate value (mental step directed to observation, evaluation – a person could determine whether to select an observed entity using an observed or determined sampled rate value in their mind);
and in response to determining to select the entity using the sampled rate value: generate a probability score for a pair of the entity and a user using the sampled rate value, wherein the probability score comprises at least one of a follow probability, a utility probability, or a create probability (mental step directed to observation, evaluation – a person could use a mentally observed or determined sampled rate value to generate a follow, utility, or create probability score for an observed entity and associated user in their mind);
determine a probability distribution for the pair using the probability score, wherein the probability distribution has a mean that is the probability score and a variance determined using a number of followers of the entity (mental step directed to observation, evaluation – a person could determine a probability distribution in their mind with an observed or mentally determined mean as a probability score and an observed or mentally determined variance using a number of followers of an entity);
sample the probability distribution to generate a sampled probability value for the pair, wherein the sampled probability value is based on the mean and the variance of the probability distribution (mental step directed to observation, evaluation – a person could sample a probability distribution in their mind to generate a sampled probability value for an observed entity and associated user based on an observed or mentally determined mean and variance);
determine a presentation list of entities of the plurality of entities based on the ranking (mental step directed to observation, evaluation – a person could determine a presentation list of entities based on a mentally observed or determined ranking in their mind).
Claim 20 recites the following additional elements:
a system comprising: at least one memory device; and a processing device; receiving a sampled rate value for an entity of a plurality of entities; train the machine learning model to rank the plurality of entities for presentation via the online system using the sampled probability value; apply the trained machine learning model to the plurality of entities; and determine, by the trained machine learning model, a ranking of the plurality of entities; and cause the presentation list of entities to be presented on the computing device.
A system comprising a memory and a processor is interpreted as a generic computer component. As the claim does not recite any particular technical details regarding the machine learning model nor any training process, the machine learning model is also interpreted as a generic computer component. Training this model, applying this model, and determining a ranking of entities using a sampled probability value are interpreted as merely using these generic computer components to apply a mental step associated with ranking entities using an observed or determined sampled probability value. Wherein the entities are ranked for presentation via the online system is interpreted as the intended use of the ranking, though Examiner notes that actively presenting ranked entities via an online system would be interpreted as transmitting data over a network. Receiving a sampled rate value for an entity and causing a presentation list of entities to be presented are interpreted as well-understood, routine conventional activity directed to transmitting and receiving data over a network.
These additional elements do not integrate the abstract idea into a practical application or amount to significantly more than the abstract idea (see MPEP 2106.05(d) and MPEP 2106.06(f)).
The independent claims are not patent eligible.
Dependent claims 2, 4-7, 9-12, 14, 16-19, and 21-22 when analyzed as a whole are held to be patent ineligible under 35 U.S.C. 101 because the additional recited limitations fail to establish that the claims are not directed to an abstract idea, as they recite further embellishment of the judicial exception.
Claim 2 recites determining whether to select the entity using the sampled rate value, wherein generating the probability score, determining the probability distribution, generating the sampled probability value, and training the machine learning model are in response to determining to select the entity using the sampled rate value (this limitation merely describes the conditions under which the mental steps (merely implemented by generic computer components) from claim 1 are performed, which does not integrate the abstract ideas into a practical application or amount to significantly more than the abstract ideas).
Claim 4 recites generating the probability score using an estimation of at least one of a follow probability, a utility probability, or a create probability, wherein the probability score comprises a weighted combination based on the at least one of the follow probability, the utility probability, or the create probability (mental step directed to observation, evaluation – a person could determine a probability score comprising a weighted combination in their mind).
Claim 5 recites applying the trained machine learning model to the plurality of entities; determining, by the trained machine learning model, a ranking of the plurality of entities; and causing a downstream action on a computing device using the ranking (As the claims do not recite any particular technical details regarding the machine learning model nor any training process, the machine learning model is also interpreted as a generic computer component. Training and applying this model to determine rankings of entities are interpreted as merely using these generic computer components to apply a mental step associated with ranking entities. Causing a downstream action using the ranking is interpreted as an additional element directed to insignificant post-solution activity, which does not integrate the claimed abstract ideas into a practical application or amount to significantly more than the abstract ideas (see MPEP 2106.05(f) and MPEP 2106.05(g)).
Claim 6 recites wherein the downstream action comprises: determining a presentation list of entities of the plurality of entities based on the ranking; and causing the presentation list of entities to be presented on the computing device (mental step directed to observation, evaluation – a person could determine a presentation list in their mind based on an observed or determined ranking. Causing a presentation list to be presented on a computing device is interpreted as well-understood, routine, conventional activity directed to transmitting data over a network, which does not integrate the claimed abstract ideas into a practical application or amount to significantly more than the abstract ideas (see MPEP 2106.05(g)).
Claim 7 recites wherein the user comprises a plurality of attributes and wherein generating the probability score comprises: determining an attribute of the plurality of attributes, wherein the attribute comprises at least one of: entity, title location, industry, or skills; and generating a probability distribution for the attribute (mental step directed to observation, evaluation – a person could determine an least an entity attribute associated with a user in their mind and generate a probability distribution for this attribute in their mind).
Claim 9 recites determining whether to select the entity using the sampled rate value, wherein generating the probability score, determining the probability distribution, generating the sampled probability value, and training the machine learning model are in response to determining to select the entity using the sampled rate value (this limitation merely describes the conditions under which the mental steps (merely implemented by generic computer components) from claim 1 are performed, which does not integrate the abstract ideas into a practical application or amount to significantly more than the abstract ideas).
Claim 10 recites applying the trained machine learning model to the plurality of entities; determining, by the trained machine learning model, a ranking of the plurality of entities; and causing a downstream action on a computing device using the ranking (As the claims do not recite any particular technical details regarding the machine learning model nor any training process, the machine learning model is also interpreted as a generic computer component. Training and applying this model to determine rankings of entities are interpreted as merely using these generic computer components to apply a mental step associated with ranking entities. Causing a downstream action using the ranking is interpreted as an additional element directed to insignificant post-solution activity, which does not integrate the claimed abstract ideas into a practical application or amount to significantly more than the abstract ideas (see MPEP 2106.05(f) and MPEP 2106.05(g)).
Claim 11 recites wherein the downstream action comprises: determining a presentation list of entities of the plurality of entities based on the ranking; and causing the presentation list of entities to be presented on the computing device (mental step directed to observation, evaluation – a person could determine a presentation list in their mind based on an observed or determined ranking. Causing a presentation list to be presented on a computing device is interpreted as well-understood, routine, conventional activity directed to transmitting data over a network, which does not integrate the claimed abstract ideas into a practical application or amount to significantly more than the abstract ideas (see MPEP 2106.05(g)).
Claim 12 recites wherein the user comprises a plurality of attributes and wherein generating the probability score comprises: determining an attribute of the plurality of attributes, wherein the attribute comprises at least one of: entity, title location, industry, or skills; and generating a probability distribution for the attribute (mental step directed to observation, evaluation – a person could determine an least an entity attribute associated with a user in their mind and generate a probability distribution for this attribute in their mind).
Claim 14 is a system claim and its limitation is included in claim 2. Claim 14 is rejected for the same reasons as claim 2.
Claim 16 is a system claim and its limitation is included in claim 4. Claim 16 is rejected for the same reasons as claim 4.
Claim 17 is a system claim and its limitation is included in claim 5. Claim 17 is rejected for the same reasons as claim 5.
Claim 18 is a system claim and its limitation is included in claim 6. Claim 18 is rejected for the same reasons as claim 6.
Claim 19 is a system claim and its limitation is included in claim 7. Claim 19 is rejected for the same reasons as claim 7.
Claim 21 is a system claim and its limitation is included in claim 9. Claim 21 is rejected for the same reasons as claim 9.
Claim 22 is a system claim and its limitation is included in claim 7. Claim 22 is rejected for the same reasons as claim 7.
Viewed as a whole, these additional claim elements do not provide meaningful limitations to transform the abstract idea into a patent eligible application of the abstract idea such that the claims amount to significantly more than the abstract idea itself. Therefore, the claims are rejected under 35 U.S.C. 101 as being directed to non-statutory subject matter.
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.
Claims 1-2, 4-14, and 16-22 are rejected under 35 U.S.C. 103 as being unpatentable over Li et al* (US 20200382530 A1, herein Li) in view of Kenthapadi et al* (US 20200272304 A1, herein Kenthapadi), in further view of Wang et al*, (“Detect Inflated Follower Numbers in OSN Using Star Sampling”, herein Wang).
*a copy of this document was provided with the office action dated 12/04/2025
*this document was included in the IDS dated 04/29/2025
Regarding claim 1, Li teaches a method (para. [0016] recites “A system and method for computing an unequal sampling probability in a highly imbalanced large population”) for training a machine learning model [for ranking entities] for presentation via an online system (para. [0043] recites “Block 250 may additionally involve causing information about each corresponding entity to be presented to an end-user” (i.e., presenting a determined list of entities to a user)), the method comprising: generating a reward score for an entity of a plurality of entities (para. [0029]-[0030] recite “FIG. 2 is a flow diagram that depicts a process for using unequal probability to sample entities from a set of entities scored by a prediction/likelihood model, in an embodiment. At block 210, a set of scores generated by a prediction model is identified, each score corresponding to a different entity of multiple entities” (i.e., a score output by a prediction model, or a reward, corresponding to an entity));
determining a rate distribution for the entity using the reward score (para. [0031] recites “At block 220, multiple buckets are determined. Each bucket corresponds to a different range of scores” (i.e., determining a range, or a distribution of each the rate of each score));
generating a sampled rate value for the entity by sampling the rate distribution (para. [0033] recites “At block 230, a probability density function (PDF) is generated based on the set of scores and the number of scores belonging to each of the buckets. A (PDF), or density of a continuous random variable, is a function whose value at any given sample (or point) in the sample space (the set of possible values taken by the random variable) can be interpreted as providing a relative likelihood that the value of the random variable would equal that sample” (i.e., generating a sample rate value for an entity from the range, or distribution of the rate of scores));
generating a probability score for a pair of the entity and a user using the sampled rate value (para. [0028] recites “An "entity" is a person or object that may be scored by likelihood model 132 based on features of the entity. Examples of entities include a user, a registered member, an account, and any user-generated content, such as an online post (e.g., a job posting), a video, an image, a (e.g., news) article, and a comment”. Para. [0038] recites “At block 240, for each score or for each score bucket, a probability of sampling that score (or a score from that bucket) based on the PDF is determined. Block 240 may involve determining a probability for each score or for each bucket” (i.e., determining a probability score for each entity, wherein the entity can be a post or account associated with a user));
determining a probability distribution for the pair using the probability score, wherein the probability distribution has a mean that is the probability score [and a variance determined using a number of followers of the entity]; generating a sampled probability value for the pair by sampling the probability distribution, wherein the sampled probability value is generated according to the mean [and the variance] of the probability distribution (para. [0039] recites “to calculate a sampling probability of a particular score, f(scorei) is calculated for all scores i that were generated by the prediction model. The inverse of each f(score) (i.e., 1/f(scorei) = si) is then computed. Then, the sum of the inverse values of all the scores (i.e., ∑sj where ∑ is from j = 1 to j = N, where N is the number of entities or scores that were generated by the prediction model) is then computed. Then, for each entity i, a sampling probability is computed for that entity based on the score for that entity: s/∑sj” (i.e., determining a probability distribution f(scorei) for a given entity and generating a sampled probability value based on the average, or mean s/∑sj for that entity)).
However, while Li teaches training a machine learning model (see at least para. [0025]) and presenting the output of the model via an online system (see at least paragraph [0043]), Li does not explicitly teach training the machine learning model to rank the plurality of entities [for presentation via the online system] based on the sampled probability value.
Kenthapadi teaches training the machine learning model to rank the plurality of entities [for presentation via the online system] based on the sampled probability value (para. [0116] recites “Ranker 202 also uses partitions 234-236 to calculate a distribution 242 of an attribute ( e.g., gender, age range, ethnicity, a combination of two or more attributes, etc.) in qualified candidates 212”. Para. [0117] recites “After a set of qualified candidates 212 matching parameters 214 is received from all partitions 234-236, ranker 202 applies a machine learning model 208 to features for qualified candidates 212 from a feature repository 238 to generate scores 216 for qualified candidates 212. Ranker 202 then produces a ranking 220 of qualified candidates 212 by scores 216” (i.e., using a trained machine learning model to determine entity rankings based on a value from a training data, or sample probability distribution)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine these teachings by utilizing the machine learning model from Kenthapadi to rank the sampled entities from Li. Li and Kenthapadi are both directed to methods of using machine learning to analyze entity distributions, and Li states in at least paragraph [0024] that “embodiments are not limited to any particular machine learning technique”. One of ordinary skill in the art would be motivated to rank the entities from Li using the model from Kenthapadi to potentially provide information regarding the entities in a more useful way to a user.
However, while Kenthapadi teaches in at least [0115] that entities can be ranked based at least in part on attributes such as number of followers, the combination of Li and Kenthapadi does not explicitly teach wherein a probability distribution has a variance determined using a number of followers of an entity.
Wang teaches wherein the probability distribution has a variance determined using a number of followers of an entity (section IV A recites “To find the follower number of the top bloggers, it is no longer effective to use uniform random sampling, where all the nodes, most of them are small ones with few connections, have equal probability of being sampled. To have top bloggers sampled more often, large nodes should have high probability of being sampled. Therefore, we opt for PPS (i.e., probability proportional to size) sampling where nodes are sampled with probability proportional to their degrees. There are several choices to run PPS sampling, such as random edge and random walk samplings. Suppose that in the directed graph G, there are N number of nodes labeled as 1, 2, . . . , N, whose in-degrees are denoted by di for i ∈ {1, 2, . . . , N}. Let E denote the number of edges, which is the sum of all the in-degrees (or out-degrees). If we select an occurrence from the sequence uniformly at random, each occurrence is selected with equal probability 1/E, and node i has the probability pi = di/E being selected. The number of times node i is selected after n sample nodes are taken can be described by the binomial distribution B(n, pi). Thus, the expected number of captures of node i is E(fi) = npi. Or, given observation of fi, pi can be estimated by (EQ3), where fi is the number of times node i is sampled. With the knowledge of pi, di, the number of followers of node i is estimated by (EQ4). Because of the binomial distribution, the variance of the estimator is (EQ5)” (i.e., determining a probability distribution with a variance based on a number of followers)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine these teachings by modifying the probability distribution sampling method from Li with the star sampling and variance estimation methods from Wang. Li and Wang are both directed to sampling probabilities from users of social computing environments. Li states in paragraph [0017] that its methods may be applied on “multiple types of entities that have an imbalanced distribution”. One of ordinary skill in the art recognize that the user network from Wang would have an imbalanced distribution, as top users would have many more followers than an average user, and be motivated to use the methods from Li in combination with Wang.
Regarding claim 2, the combination of Li, Kenthapadi, and Wang teaches the method of claim 1, further comprising: determining whether to select the entity using the sampled rate value, wherein generating the probability score, determining the probability distribution, generating the sampled probability value, and training the machine learning model are in response to determining to select the entity using the sampled rate value (Li para. [0035] recites “block 230 may comprise testing different values of one or more parameters of a PDF function relative to the score distribution until a loss is minimized or until the loss is below some pre-defined threshold” (i.e., using the sampled rate value determined by the probability density function to select a given entity)).
Regarding claim 4, the combination of Li, Kenthapadi, and Wang teaches the method of claim 1, further comprising generating the probability score using an estimation of at least one of a follow probability, a utility probability, or a create probability, wherein the probability score comprises a weighted combination based on the at least one of the follow probability, the utility probability, or the create probability (Li para. [0046] recites “weights are determined based on activity levels and used to adjust the sampling probability. In the context of entities as accounts, multiple activities may be tracked, such as number of messages transmitted from the account, number of posts uploaded from the account, number of advertisements provided by a user of the account (if the account corresponds to a content/advertisement provider), number of advertisements selected by a user of the account, and number of content item (e.g., advertisement) impressions by a user of the account. In the context of entities as posts, multiple activities may be tracked, such as number of user feeds the post has appeared (e.g., number of impressions), number of times other users have selected the post (e.g., number of clicks), number of negative interactions of the post (e.g., number of down votes and/or number of negative comments), etc.” (i.e., the probability score can be determined by at least a weighted combination of factors including a usage, or utility probability)).
Regarding claim 5, the combination of Li, Kenthapadi, and Wang teaches the method of claim 1, further comprising: applying the trained machine learning model to the plurality of entities; determining, by the trained machine learning model, a ranking of the plurality of entities (Kenthapadi para. [0117] recites “After a set of qualified candidates 212 matching parameters 214 is received from all partitions 234-236, ranker 202 applies a machine learning model 208 to features for qualified candidates 212 from a feature repository 238 to generate scores 216 for qualified candidates 212. Ranker 202 then produces a ranking 220 of qualified candidates 212 by scores 216” (i.e., using a trained machine learning model to determine entity rankings));
and causing a downstream action on a computing device using the ranking (Kenthapadi para. [0118] recites “Ranker 202 then uses distribution 242 to perform a reranking 224 of qualified candidates 212 to reduce bias from machine learning model 208 in generating ranking 220” (i.e., causing a reranking, or downstream action, using the initial ranking)).
Regarding claim 6, the combination of Li, Kenthapadi, and Wang teaches the method of claim 5, wherein the downstream action comprises: determining a presentation list of entities of the plurality of entities based on the ranking; and causing the presentation list of entities to be presented on the computing device (Kenthapadi para. [0121] recites “Finally, management apparatus 206 outputs some or all of reranking 226 in a result 228 for the request. For example, management apparatus 206 may paginate some or all candidates in reranking 226 into subsets of result 228 that are displayed as a user scrolls through entries in result 228 and/or navigates across screens or pages containing result 228”. Li para. [0043] recites “Block 250 may additionally involve causing information about each corresponding entity to be presented to an end-user” (i.e., presenting a determined ranked list of entities to a user)).
Regarding claim 7, the combination of Li, Kenthapadi, and Wang teaches the method of claim 1, wherein the user comprises a plurality of attributes and wherein generating the probability score comprises: determining an attribute of the plurality of attributes, wherein the attribute comprises at least one of: entity, title location, industry, or skills (Li para. [0023] recites “a machine learning technique is used to generate a statistical model that is trained based on a history of attribute values associated with users. The statistical model is trained based on multiple attributes. In machine learning parlance, such attributes are referred to as "features"”. Li para. [0025] recites “Example user-related features includes job title, industry, job function, employer, academic degrees, geographical location, skills” (i.e., determining attributes associated with an entity including at least location, industry, or skills)); and generating a probability distribution for the attribute (Li para. [0026] recites “In order to generate the training data, information about each entity is analyzed to compute the different feature values”. Li para. [0030] recites “At block 210, a set of scores generated by a prediction model is identified, each score corresponding to a different entity of multiple entities”. Li para. [0032] recites “FIG. 3 is a histogram that depicts an example score distribution (of scores generated by a likelihood model) that is simulated with a Weibull distribution” (i.e., generating a probability distribution for the features, or attributes, of a given entity)).
Regarding claim 8, Li teaches a method (para. [0016] recites “A system and method for computing an unequal sampling probability in a highly imbalanced large population”) for training a machine learning model [for ranking] for presentation via an online system (para. [0043] recites “Block 250 may additionally involve causing information about each corresponding entity to be presented to an end-user” (i.e., presenting a determined list of entities to a user)), the method comprising: receiving a sampled rate value for an entity of a plurality of entities (para. [0033] recites “At block 230, a probability density function (PDF) is generated based on the set of scores and the number of scores belonging to each of the buckets. A (PDF), or density of a continuous random variable, is a function whose value at any given sample (or point) in the sample space (the set of possible values taken by the random variable) can be interpreted as providing a relative likelihood that the value of the random variable would equal that sample” (i.e., receiving a generated sample rate value for an entity from the range, or distribution of the rate of scores));
generating a probability score for a pair of the entity and a user using the sampled rate value, wherein the probability score comprises at least one of a follow probability, a utility probability, or a create probability (Li para. [0033] recites “At block 230, a probability density function (PDF) is generated based on the set of scores and the number of scores belonging to each of the buckets. A (PDF), or density of a continuous random variable, is a function whose value at any given sample (or point) in the sample space (the set of possible values taken by the random variable) can be interpreted as providing a relative likelihood that the value of the random variable would equal that sample” (i.e., generating a sample rate value for an entity from the range, or distribution of the rate of scores)). Li para. [0046] recites “weights are determined based on activity levels and used to adjust the sampling probability. In the context of entities as accounts, multiple activities may be tracked, such as number of messages transmitted from the account, number of posts uploaded from the account, number of advertisements provided by a user of the account (if the account corresponds to a content/advertisement provider), number of advertisements selected by a user of the account, and number of content item (e.g., advertisement) impressions by a user of the account. In the context of entities as posts, multiple activities may be tracked, such as number of user feeds the post has appeared (e.g., number of impressions), number of times other users have selected the post (e.g., number of clicks), number of negative interactions of the post (e.g., number of down votes and/or number of negative comments), etc.” (i.e., the probability score can be determined by at least a weighted combination of factors including a usage, or utility probability));
determining a probability distribution for the pair using the probability score, wherein the probability distribution has a mean that is the probability score [and a variance determined using a number of followers of the entity]; sampling the probability distribution to generate a sampled probability value for the pair, wherein the sampled probability value is based on the mean [and the variance] of the probability distribution (para. [0039] recites “to calculate a sampling probability of a particular score, f(scorei) is calculated for all scores i that were generated by the prediction model. The inverse of each f(score) (i.e., 1/f(scorei) = si) is then computed. Then, the sum of the inverse values of all the scores (i.e., ∑sj where ∑ is from j = 1 to j = N, where N is the number of entities or scores that were generated by the prediction model) is then computed. Then, for each entity i, a sampling probability is computed for that entity based on the score for that entity: s/∑sj” (i.e., determining a probability distribution f(scorei) for a given entity and generating a sampled probability value based on the average, or mean s/∑sj for that entity)).
However, while Li teaches training a machine learning model (see at least para. [0025]) and presenting the output of the model via an online system (see at least paragraph [0043]), Li does not explicitly teach training the machine learning model to rank the plurality of entities [for presentation via the online system] based on the sampled probability value.
Kenthapadi teaches training the machine learning model to rank the plurality of entities [for presentation via the online system] based on the sampled probability value (Kenthapadi para. [0117] recites “After a set of qualified candidates 212 matching parameters 214 is received from all partitions 234-236, ranker 202 applies a machine learning model 208 to features for qualified candidates 212 from a feature repository 238 to generate scores 216 for qualified candidates 212. Ranker 202 then produces a ranking 220 of qualified candidates 212 by scores 216” (i.e., using a trained machine learning model to determine entity rankings)).
See claim 1 for motivation to combine.
However, while Kenthapadi teaches in at least [0115] that entities can be ranked based at least in part on attributes such as number of followers, the combination of Li and Kenthapadi does not explicitly teach wherein a probability distribution has a variance determined using a number of followers of an entity.
Wang teaches wherein the probability distribution has a variance determined using a number of followers of an entity (section IV A recites “To find the follower number of the top bloggers, it is no longer effective to use uniform random sampling, where all the nodes, most of them are small ones with few connections, have equal probability of being sampled. To have top bloggers sampled more often, large nodes should have high probability of being sampled. Therefore, we opt for PPS (i.e., probability proportional to size) sampling where nodes are sampled with probability proportional to their degrees. There are several choices to run PPS sampling, such as random edge and random walk samplings. Suppose that in the directed graph G, there are N number of nodes labeled as 1, 2, . . . , N, whose in-degrees are denoted by di for i ∈ {1, 2, . . . , N}. Let E denote the number of edges, which is the sum of all the in-degrees (or out-degrees). If we select an occurrence from the sequence uniformly at random, each occurrence is selected with equal probability 1/E, and node i has the probability pi = di/E being selected. The number of times node i is selected after n sample nodes are taken can be described by the binomial distribution B(n, pi). Thus, the expected number of captures of node i is E(fi) = npi. Or, given observation of fi, pi can be estimated by (EQ3), where fi is the number of times node i is sampled. With the knowledge of pi, di, the number of followers of node i is estimated by (EQ4). Because of the binomial distribution, the variance of the estimator is (EQ5)” (i.e., determining a probability distribution with a variance based on a number of followers)).
See claim 1 for motivation to combine.
Regarding claim 9, the combination of Li, Kenthapadi, and Wang teaches the method of claim 8, further comprising: determining whether to select the entity using the sampled rate value, wherein generating the probability score, determining the probability distribution, generating the sampled probability value, and training the machine learning model are in response to determining to select the entity using the sampled rate value (Li para. [0035] recites “block 230 may comprise testing different values of one or more parameters of a PDF function relative to the score distribution until a loss is minimized or until the loss is below some pre-defined threshold” (i.e., using the sampled rate value determined by the probability density function to select a given entity)).
Regarding claim 10, the combination of Li, Kenthapadi, and Wang teaches the method of claim 8, further comprising: applying the trained machine learning model to the plurality of entities; determining, by the trained machine learning model, a ranking of the plurality of entities (Kenthapadi para. [0117] recites “After a set of qualified candidates 212 matching parameters 214 is received from all partitions 234-236, ranker 202 applies a machine learning model 208 to features for qualified candidates 212 from a feature repository 238 to generate scores 216 for qualified candidates 212. Ranker 202 then produces a ranking 220 of qualified candidates 212 by scores 216” (i.e., applying a trained machine learning model to determine entity rankings));
and causing a downstream action on a computing device using the ranking (Kenthapadi para. [0118] recites “Ranker 202 then uses distribution 242 to perform a reranking 224 of qualified candidates 212 to reduce bias from machine learning model 208 in generating ranking 220” (i.e., causing a reranking, or downstream action, using the initial ranking)).
Regarding claim 11, the combination of Li, Kenthapadi, and Wang teaches the method of claim 10, wherein the downstream action comprises: determining a presentation list of entities of the plurality of entities based on the ranking; and causing the presentation list of entities to be presented on the computing device (Kenthapadi para. [0121] recites “Finally, management apparatus 206 outputs some or all of reranking 226 in a result 228 for the request. For example, management apparatus 206 may paginate some or all candidates in reranking 226 into subsets of result 228 that are displayed as a user scrolls through entries in result 228 and/or navigates across screens or pages containing result 228”. Li para. [0043] recites “Block 250 may additionally involve causing information about each corresponding entity to be presented to an end-user” (i.e., presenting a determined ranked list of entities to a user)).
Regarding claim 12, the combination of Li, Kenthapadi, and Wang teaches the method of claim 8, wherein the user comprises a plurality of attributes and wherein generating the probability score comprises: determining an attribute of the plurality of attributes, wherein the attribute comprises at least one of: entity, title location, industry, or skills (Li para. [0023] recites “a machine learning technique is used to generate a statistical model that is trained based on a history of attribute values associated with users. The statistical model is trained based on multiple attributes. In machine learning parlance, such attributes are referred to as "features"”. Li para. [0025] recites “Example user-related features includes job title, industry, job function, employer, academic degrees, geographical location, skills” (i.e., determining attributes associated with an entity including at least location, industry, or skills)); and generating a probability distribution for the attribute (Li para. [0026] recites “In order to generate the training data, information about each entity is analyzed to compute the different feature values”. Li para. [0030] recites “At block 210, a set of scores generated by a prediction model is identified, each score corresponding to a different entity of multiple entities”. Li para. [0032] recites “FIG. 3 is a histogram that depicts an example score distribution (of scores generated by a likelihood model) that is simulated with a Weibull distribution” (i.e., generating a probability distribution for the features, or attributes, of a given entity)).
Claim 13 is a system claim and its limitation is included in claim 1. The only difference is that claim 13 requires a system (Li para. [0016] recites “A system and method for computing an unequal sampling probability in a highly imbalanced large population”). Therefore, claim 13 is rejected for the same reasons as claim 1.
Claim 14 is a system claim and its limitation is included in claim 2. Claim 14 is rejected for the same reasons as claim 2.
Claim 16 is a system claim and its limitation is included in claim 4. Claim 16 is rejected for the same reasons as claim 4.
Claim 17 is a system claim and its limitation is included in claim 5. Claim 17 is rejected for the same reasons as claim 5.
Claim 18 is a system claim and its limitation is included in claim 6. Claim 18 is rejected for the same reasons as claim 6.
Claim 19 is a system claim and its limitation is included in claim 7. Claim 19 is rejected for the same reasons as claim 7.
Regarding claim 20, Li teaches a system for training a machine learning model [to rank entities] for presentation via an online system (para. [0043] recites “Block 250 may additionally involve causing information about each corresponding entity to be presented to an end-user” (i.e., presenting a determined list of entities to a user)), the system comprising: at least one memory device; and a processing device, operatively coupled with the at least one memory device (para. [0058] recites “Computer system 800 further includes a read only memory (ROM) 808 or other static storage device coupled to bus 802 for storing static information and instructions for processor 804”), to:
receive a sampled rate value for an entity of a plurality of entities (para. [0033] recites “At block 230, a probability density function (PDF) is generated based on the set of scores and the number of scores belonging to each of the buckets. A (PDF), or density of a continuous random variable, is a function whose value at any given sample (or point) in the sample space (the set of possible values taken by the random variable) can be interpreted as providing a relative likelihood that the value of the random variable would equal that sample” (i.e., receiving a generated sample rate value for an entity from the range, or distribution of the rate of scores));
determine whether to select the entity using the sampled rate value (para. [0035] recites “block 230 may comprise testing different values of one or more parameters of a PDF function relative to the score distribution until a loss is minimized or until the loss is below some pre-defined threshold” (i.e., using the sampled rate value determined by the probability density function to select a given entity that meets a given condition));
and in response to determining to select the entity using the sampled rate value: generate a probability score for a pair of the entity and a user using the sampled rate value, wherein the probability score comprises at least one of a follow probability, a utility probability, or a create probability (para. [0046] recites “weights are determined based on activity levels and used to adjust the sampling probability. In the context of entities as accounts, multiple activities may be tracked, such as number of messages transmitted from the account, number of posts uploaded from the account, number of advertisements provided by a user of the account (if the account corresponds to a content/advertisement provider), number of advertisements selected by a user of the account, and number of content item (e.g., advertisement) impressions by a user of the account. In the context of entities as posts, multiple activities may be tracked, such as number of user feeds the post has appeared (e.g., number of impressions), number of times other users have selected the post (e.g., number of clicks), number of negative interactions of the post (e.g., number of down votes and/or number of negative comments), etc.” (i.e., the probability score can be determined by at least a weighted combination of factors including a usage, or utility probability));
determining a probability distribution for the pair using the probability score, wherein the probability distribution has a mean that is the probability score [and a variance determined using a number of followers of the entity]; sampling the probability distribution to generate a sampled probability value for the pair, wherein the sampled probability value is based on the mean [and the variance] of the probability distribution (para. [0039] recites “to calculate a sampling probability of a particular score, f(scorei) is calculated for all scores i that were generated by the prediction model. The inverse of each f(score) (i.e., 1/f(scorei) = si) is then computed. Then, the sum of the inverse values of all the scores (i.e., ∑sj where ∑ is from j = 1 to j = N, where N is the number of entities or scores that were generated by the prediction model) is then computed. Then, for each entity i, a sampling probability is computed for that entity based on the score for that entity: s/∑sj” (i.e., determining a probability distribution f(scorei) for a given entity and generating a sampled probability value based on the average, or mean s/∑sj for that entity));
determine a presentation list of entities of the plurality of entities [based on the ranking]; and cause the presentation list of entities to be presented on the computing device (Li para. [0043] recites “Block 250 may additionally involve causing information about each corresponding entity to be presented to an end-user” (i.e., presenting a determined list of entities to a user)).
However, while Li teaches training a machine learning model (see at least para. [0025]) and presenting the output of the model via an online system (see at least paragraph [0043]), Li does not explicitly teach training the machine learning model to rank the plurality of entities [for presentation via the online system] based on the sampled probability value; applying the trained machine learning model to the plurality of entities; and determine, by the trained machine learning model, a ranking of the plurality of entities.
Kenthapadi teaches training the machine learning model to rank the plurality of entities [for presentation via the online system] based on the sampled probability value; apply the trained machine learning model to the plurality of entities; and determine, by the trained machine learning model, a ranking of the plurality of entities (Kenthapadi para. [0117] recites “After a set of qualified candidates 212 matching parameters 214 is received from all partitions 234-236, ranker 202 applies a machine learning model 208 to features for qualified candidates 212 from a feature repository 238 to generate scores 216 for qualified candidates 212. Ranker 202 then produces a ranking 220 of qualified candidates 212 by scores 216” (i.e., using a trained machine learning model to determine entity rankings)).
See claim 1 for motivation to combine.
However, while Kenthapadi teaches in at least [0115] that entities can be ranked based at least in part on attributes such as number of followers, the combination of Li and Kenthapadi does not explicitly teach wherein a probability distribution has a variance determined using a number of followers of an entity.
Wang teaches wherein the probability distribution has a variance determined using a number of followers of an entity (section IV A recites “To find the follower number of the top bloggers, it is no longer effective to use uniform random sampling, where all the nodes, most of them are small ones with few connections, have equal probability of being sampled. To have top bloggers sampled more often, large nodes should have high probability of being sampled. Therefore, we opt for PPS (i.e., probability proportional to size) sampling where nodes are sampled with probability proportional to their degrees. There are several choices to run PPS sampling, such as random edge and random walk samplings. Suppose that in the directed graph G, there are N number of nodes labeled as 1, 2, . . . , N, whose in-degrees are denoted by di for i ∈ {1, 2, . . . , N}. Let E denote the number of edges, which is the sum of all the in-degrees (or out-degrees). If we select an occurrence from the sequence uniformly at random, each occurrence is selected with equal probability 1/E, and node i has the probability pi = di/E being selected. The number of times node i is selected after n sample nodes are taken can be described by the binomial distribution B(n, pi). Thus, the expected number of captures of node i is E(fi) = npi. Or, given observation of fi, pi can be estimated by (EQ3), where fi is the number of times node i is sampled. With the knowledge of pi, di, the number of followers of node i is estimated by (EQ4). Because of the binomial distribution, the variance of the estimator is (EQ5)” (i.e., determining a probability distribution with a variance based on a number of followers)).
See claim 1 for motivation to combine.
Claim 21 is a system claim and its limitation is included in claim 9. Claim 21 is rejected for the same reasons as claim 9.
Claim 22 is a system claim and its limitation is included in claim 7. Claim 22 is rejected for the same reasons as claim 7.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US 11838309 B1 (Martin et al) teaches a method for analyzing social media accounts, including monitoring and analyzing the variance in a number of followers for a given account.
US 20240078411 A1 (Beye et al) teaches a method for determining a probability distribution using a Gaussian Mixture Model based on a parameter set including a weight coefficient, mean, and variance.
“Thompson Sampling Algorithms for Mean-Variance Bandits” (Zhu et al) teaches a method for Thompson Sampling-style algorithms for mean-variance multi-armed bandits and provide comprehensive regret analyses for Gaussian and Bernoulli bandits.
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 LEAH M FEITL whose telephone number is (571) 272-8350. The examiner can normally be reached on M-F 0900-1700 EST.
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, Viker Lamardo can be reached on (571) 270-5871. 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.
/L.M.F./ Examiner, Art Unit 2147
/ERIC NILSSON/ Primary Examiner, Art Unit 2151
Prosecution Insights