DETAILED ACTION
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination
2. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant’s submission filed 02 July 2026 [hereinafter Response], where:
Claims 1, 3, 4-6, 8, 9, 11-14, 16, 17, 19, and 20 have been amended.
Claims 7 and 15 have been cancelled.
Claims 1-6, 8-14, and 16-20 are pending.
Claims 1-6, 8-14, and 16-20 are rejected.
Claim Rejections - 35 U.S.C. § 101
3. 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.
4. Claims 1-6, 8-14, and 16-20 are rejected under 35 U.S.C. § 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites a system, which is a machine, and thus one of the statutory categories of patentable subject matter. (35 U.S.C. § 101).
However, under Step 2A Prong One, the claim recites the limitations of “generate, using a representation learning, a representation space comprising vector representations of cardholders and merchants based on relationships learned from transaction history,” “extract groups of representations from the representation space by grouping together representations corresponding to a same merchant category code (MCC),” “generate predictions including future times of future transactions involving a user account, and specific merchants expected to coincide with the future times based on the probability density function of the LSTM RNN,” and “filter irrelevant information including spam from the user based on the prediction of the probability density function.”
The activities of “generate a representation space, “extract groups of representations,” “generate predictions,” and “filter irrelevant information,” include limitations that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, are mental processes, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)). Thus, claim 1 recites an abstract idea.
Under Step 2A Prong Two, the claim as a whole is not integrated into a practical application, because the additional elements recited in the claim beyond the identified judicial exception include “memory configured to store instructions of a predictor model,” and “transaction manager comprising a processor configured to execute the instructions,” which are recited at a high-level of generality, and are generic computer components used to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application. The claim also recites a “predictor model” and a “LSTM-RNN,” which are also recited at a high level of generality, and thus are generic computer components used to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application.
The claim recites the limitation of “train a long short-term memory recurrent neural network (LSTM RNN) to learn transaction patterns of accounts corresponding to different merchant category codes (MCCs) and transaction amounts by feeding, in temporal order, at each time step, transactions associated with merchants and cardholders, the merchants and contexts of the merchants and the cardholders to the LSTM RNN,” and “train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins;” where the additional element of “train” is an activity of using the generic computer components (LSTM-RNN) to implement the abstract idea, that does not serve to integrate the abstract idea into a practical application. (MPEP § 2106.05(f)).
The claim also recites more details or specifics to the additional element of “train to learn transaction patterns,” “wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,” and “the contexts are vectors representing encoded parameters including parameter a corresponding to a category class, parameter b corresponding to a class of progress percentage, parameter c corresponding to a class of proximity, and category d corresponding to a class of historical redemption,” and accordingly, are merely more specific to the additional element.
The claim also recites “automatically transmit a benefit to an owner of the user account based on the predicted future times and the specific merchants,” which is a post-processing, insignificant extra solution activity of result transmission, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. Therefore, claim 1 is directed to the abstract idea.
Finally, under Step 2B, the additional elements, taken alone or in combination, do not represent significantly more than the abstract idea itself. The additional elements recited in the claim beyond the identified judicial exception include “memory configured to store instructions of a predictor model,” and “transaction manager comprising a processor configured to execute the instructions,” which are recited at a high-level of generality, and are generic computer components used to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea. The claim also recites a “predictor model” and a “LSTM-RNN,” which are also recited at a high level of generality, and thus are generic computer components used to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea.
The claim recites the limitation of “train a long short-term memory recurrent neural network (LSTM RNN) to learn transaction patterns of accounts corresponding to different merchant category codes (MCCs) and transaction amounts by feeding, in temporal order, at each time step, transactions associated with merchants and cardholders, the merchants and contexts of the merchants and the cardholders to the LSTM RNN,” and “train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins;” where the additional element of “train” is an activity of using the generic computer components (LSTM-RNN) to implement the abstract idea, that does not amount to significantly more than the abstract idea. (MPEP § 2106.05(f)).
The claim also recites more details or specifics to the additional element of “train to learn transaction patterns,” “wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,” and “the contexts are vectors representing encoded parameters including parameter a corresponding to a category class, parameter b corresponding to a class of progress percentage, parameter c corresponding to a class of proximity, and category d corresponding to a class of historical redemption,” and accordingly, are merely more specific to the additional element.
The claim also recites “automatically transmit a benefit to an owner of the user account based on the predicted future times and the specific merchants,” which is a well-understood, routine, and conventional activity of receiving or transmitting data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. Thus, claim 1 is subject-matter ineligible.
Claim 9 recites a method, which is a process, and thus one of the statutory categories of patentable subject matter. (35 U.S.C. § 101).
However, under Step 2A Prong One, the claim recites the limitations of “generating, using a representation learning, a representation space comprising vector representations of cardholders and merchants based on relationships learned from transaction history,” “extracting groups of representations from the representation space by grouping together representations corresponding to a same merchant category code (MCC),” “generating predictions including future times of future transactions involving a user account, and specific merchants expected to coincide with the future times based on the probability density function of the LSTM RNN,” and “filtering irrelevant information including spam from the user based on the prediction of the probability density function.”
The activities of “generating a representation space, “extracting groups of representations,” “generating predictions,” and “filtering irrelevant information,” include limitations that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, are mental processes, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)). Thus, claim 9 recites an abstract idea.
Under Step 2A Prong Two, the claim as a whole is not integrated into a practical application, because the additional elements recited in the claim beyond the identified judicial exception include a “LSTM-RNN” which is recited at a high level of generality, and thus is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application.
The claim recites “training a long short-term memory recurrent neural network (LSTM RNN) to learn transaction patterns of accounts corresponding to different merchant category codes (MCCs) and transaction amounts by feeding, in temporal order, at each time step, transactions associated with merchants and cardholders, the merchants and contexts of the merchants and the cardholders to the LSTM RNN,” and “train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins;” where the additional element of “training” is an activity of using the generic computer components (LSTM-RNN) to implement the abstract idea, that does not serve to integrate the abstract idea into a practical application. (MPEP § 2106.05(f)).
The claim recites more details or specifics to the additional element of “training to learn transaction patterns,” “wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,” and “the contexts are vectors representing encoded parameters including parameter a corresponding to a category class, parameter b corresponding to a class of progress percentage, parameter c corresponding to a class of proximity, and category d corresponding to a class of historical redemption,” and accordingly, are merely more specific to the additional element.
The claim also recites the limitation of “automatically transmitting a benefit to an owner of the user account based on the predicted future times and the specific merchants,” which is a post-processing, insignificant extra solution activity of result transmission, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. Therefore, claim 9 is directed to the abstract idea.
Finally, under Step 2B, the additional elements, taken alone or in combination, do not represent significantly more than the abstract idea itself. The additional elements recited in the claim beyond the identified judicial exception include a “LSTM-RNN” which is recited at a high level of generality, and thus are generic computer components used to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea.
The claim recites “training a long short-term memory recurrent neural network (LSTM RNN) to learn transaction patterns of accounts corresponding to different merchant category codes (MCCs) and transaction amounts by feeding, in temporal order, at each time step, transactions associated with merchants and cardholders, the merchants and contexts of the merchants and the cardholders to the LSTM RNN,” and “train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins;”,” where the additional element of “training” is an activity of using the generic computer components (LSTM-RNN) to implement the abstract idea, that does not amount to significantly more than the abstract idea. (MPEP § 2106.05(f)).
The claim recites more details or specifics to the additional element of “training to learn transaction patterns,” “wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,” and “the contexts are vectors representing encoded parameters including parameter a corresponding to a category class, parameter b corresponding to a class of progress percentage, parameter c corresponding to a class of proximity, and category d corresponding to a class of historical redemption,” and accordingly, are merely more specific to the additional element.
The claim also recites the limitation of “automatically transmitting a benefit to an owner of the user account based on the predicted future times and the specific merchants,” which is a well-understood, routine, and conventional activity of receiving or transmitting data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. Thus, claim 9 is subject-matter ineligible.
Claim 17 recites a non-transitory, computer-readable media, which is a product, and thus one of the statutory categories of patentable subject matter. (35 U.S.C. § 101).
However, under Step 2A Prong One, the claim recites the limitations of “generating, using a representation learning, a representation space comprising vector representations of cardholders and merchants based on relationships learned from transaction history,” “extracting groups of representations from the representation space by grouping together representations corresponding to a same merchant category code (MCC),” “generating predictions including future times of future transactions involving a user account, and specific merchants expected to coincide with the future times based on the probability density function of the LSTM RNN,” and “filtering irrelevant information including spam from the user based on the prediction of the probability density function.”
The activities of “generating a representation space, “extracting groups of representations,” “generating predictions,” and “filtering irrelevant information,” include limitations that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, are mental processes, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)). Thus, claim 17 recites an abstract idea.
Under Step 2A Prong Two, the claim as a whole is not integrated into a practical application, because the additional elements recited in the claim beyond the identified judicial exception include “non-transitory, computer-readable media having computer-readable instructions stored thereon, the computer-readable instructions being capable of being read by a transaction manager configured to execute instructions stored on a memory,” which are recited at a high-level of generality, and are generic computer components used to implement the abstract idea. (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application. The claim also recites a “LSTM-RNN” which is recited at a high level of generality, and thus is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not serve to integrate the abstract idea into a practical application.
The claim recites “training a long short-term memory recurrent neural network (LSTM RNN) to learn transaction patterns of accounts corresponding to different merchant category codes (MCCs) and transaction amounts by feeding, in temporal order, at each time step, transactions associated with merchants and cardholders, the merchants and contexts of the merchants and the cardholders to the LSTM RNN,” and “train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins;” where the additional element of “training” is an activity of using the generic computer components (LSTM-RNN) to implement the abstract idea, that does not serve to integrate the abstract idea into a practical application. (MPEP § 2106.05(f)).
The claim recites more details or specifics to the additional element of “training to learn transaction patterns,” “wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,” and “the contexts are vectors representing encoded parameters including parameter a corresponding to a category class, parameter b corresponding to a class of progress percentage, parameter c corresponding to a class of proximity, and category d corresponding to a class of historical redemption,” and accordingly, are merely more specific to the additional element.
The claim recites the limitation of “automatically transmitting a benefit to an owner of the user account based on the predicted future times and the specific merchants,” which is a post-processing, insignificant extra solution activity of result transmission, (MPEP § 2106.05(g)), that does not serve to integrate the abstract idea into a practical application. Therefore, claim 17 is directed to the abstract idea.
Finally, under Step 2B, the additional elements, taken alone or in combination, do not represent significantly more than the abstract idea itself. The additional elements recited in the claim beyond the identified judicial exception include “non-transitory, computer-readable media having computer-readable instructions stored thereon, the computer-readable instructions being capable of being read by a transaction manager configured to execute instructions stored on a memory,” which are recited at a high-level of generality, and are generic computer components used to implement the abstract idea. (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea. The claim also recites a “LSTM-RNN” which is recited at a high level of generality, and thus is a generic computer component used to implement the abstract idea, (MPEP § 2106.05(f)), that does not amount to significantly more than the abstract idea.
The claim recites “training a long short-term memory recurrent neural network (LSTM RNN) to learn transaction patterns of accounts corresponding to different merchant category codes (MCCs) and transaction amounts by feeding, in temporal order, at each time step, transactions associated with merchants and cardholders, the merchants and contexts of the merchants and the cardholders to the LSTM RNN,” and “train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins;” where the additional element of “training” is an activity of using the generic computer components (LSTM-RNN) to implement the abstract idea, that does not amount to significantly more than the abstract idea. (MPEP § 2106.05(f)).
The claim recites more details or specifics to the additional element of “training to learn transaction patterns,” “wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,” and “the contexts are vectors representing encoded parameters including parameter a corresponding to a category class, parameter b corresponding to a class of progress percentage, parameter c corresponding to a class of proximity, and category d corresponding to a class of historical redemption,” and accordingly, are merely more specific to the additional element.
The claim recites the limitation of “automatically transmitting a benefit to an owner of the user account based on the predicted future times and the specific merchants which is a well-understood, routine, and conventional activity of receiving or transmitting data over a network, (MPEP § 2106.05(d) sub II.i), that does not amount to significantly more than the abstract idea. Thus, claim 17 is subject-matter ineligible.
Claim 2 depends from claim 1. Claim 10 depends from claim 9. Claim 18 depends from claim 17. The claims recite more details or specifics to the additional element of “training,” (claims 2, 10, and 18: “learning a distribution of an expenditure pattern relating the account over time”), and accordingly is merely more specific to the abstract idea. Thus, claims 2, 10, and 18 are subject-matter ineligible.
Claim 3 depends directly or indirectly from claim 1. Claim 11 depends directly or indirectly from claim 9. Claim 19 depends directly or indirectly from claim 17. The claims recite more details or specifics to the additional element of training, (claims 3, 11, and 19: providing information related to a cardholder of the account transacting at time t; providing information related to a merchant being transacted at by the cardholder of the user account at the time t; and providing a context of the cardholder and the merchant at time t), and accordingly, are merely more specific to the abstract idea. Thus, claims 3, 11, and 19 are subject-matter ineligible.
Claim 4 depends directly or indirectly from claim 1. Claim 12 depends directly or indirectly from claim 9. Claim 20 depends directly or indirectly from claim 17. The claims recite more details or specifics to the abstract idea of “predicting” (claims 4, 12, and 20: predicting transactions behavior of the account during a period”) and accordingly, are merely more specific to the abstract idea. Thus, claims 4, 12, and 20 are subject-matter ineligible.
Claim 5 depends directly or indirectly from claim 1. Claim 13 depends directly or indirectly from claim 9. The claims recite more details or specifics to the abstract idea of “predicting,” (claims 5 and 13: linking the account to a vector attribute of the future transaction involving the user account”), and accordingly, are merely more specific to the abstract idea. Therefore, claims 5 and 13 are subject-matter ineligible.
Claim 6 depends directly or indirectly from claim 1. Claim 14 depends directly or indirectly from claim 9. The claims recite more details or specifics to the abstract idea of “predicting,” (claims 6 and 14: “wherein the benefit includes at least one of a discount, offer, conditional reward, or incentive of a merchant”), and accordingly, are merely more specific to the abstract idea. Therefore, claims 6 and 14 are subject-matter ineligible.
Claim 8 depends directly or indirectly from claim 1. Claim 16 depends directly or indirectly from claim 9. The claims recite more details or specifics to the abstract idea of “predicting,” (claims 8 and 16: the benefit includes an existing or future offer provided by a merchant in the MCCs), and accordingly, is merely more specific to the abstract idea. Therefore, claims 8, and 16 are subject-matter ineligible.
Claim Rejections – 35 U.S.C. § 103
5. 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.
6. The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. § 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
7. This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. § 102(b)(2)(C) for any potential 35 U.S.C. § 102(a)(2) prior art against the later invention.
8. Claims 1-6, 8-14, and 16-20 are rejected under 35 U.S.C. § 103 as being unpatentable over Manzoor et al., “RUSH! Targeted Time-Limited Coupons via Purchase Forecasts,” KDD (2017) [hereinafter Manzoor] in view of US Published Application 20190278378 to Yan et al. [hereinafter Yan].
Regarding claims 1, 9, and 17, Manzoor teaches [a] system for managing a payment network (Manzoor, right column of p. 1923, “1. Introduction,” first partial paragraph, teaches “we partner with a national bank that provides us with a large, anonymized database that comprises transactions from 200,000 customers over a period from September 2013 to January 2016 [(that is, a system for managing a payment network)]”) of claim 1, [a] method for managing a payment network (Manzoor, abstract, teaches “a method that can predict both event time and category, adapt to temporal dynamics, handle event sparsity, that is also scalable and interpretable) of claim 9, and [a] non-transitory, computer-readable media . . . capable of instructing the transaction manager (Manzoor, left column of p. 1930, “5. Model Evaluation,” first paragraph, teaches “Scalability. . . . Parameter inference of the proposed model is linear in number of events, as shown empirically in Fig. 2. The runtime is around 60 minutes to fit RUSH! to ~2.8 million transactions. All experiments were performed on an Intel Xeon E7-8860 v3 at 2.2Ghz with 4 physical CPUS and 4 cores per CPU [(that is, the platform inherently includes a non-transitory, computer-readable media . . . capable of instructing the transaction manager)]”) of claim 17, comprising:
a memory configured to store instructions of a predictor model; and a transaction manager comprising a processor (Manzoor, left column of p. 1930, “5. Model Evaluation,” first paragraph, teaches “Scalability. . . . Parameter inference of the proposed model is linear in number of events, as shown empirically in Fig. 2. The runtime is around 60 minutes to fit RUSH! to ~2.8 million transactions. All experiments were performed on an Intel Xeon E7-8860 v3 at 2.2Ghz with 4 physical CPUS and 4 cores per CPU [(that is, inherently, a memory configured to store instructions of a predictor model; and a transaction manager comprising a processor)]”) configured to execute the instructions to:
generate, using a representation learner, a representation space comprising vector representations of cardholders and merchants based on relationships learned from transaction history (Manzoor, right column of p. 1924, “2. Problem Overview,“ first paragraph, teaches “[l]et [T0;T) be the window of observation wherein we observe all the transactions of every consumer [(that is, based on relationships learned from transaction history)]. We may assume T0 = 0 without loss of generality. Each consumer u [(that is, cardholders)] is represented by a continuous-time sequence of transaction timestamps (interchangeably called events) Tu = {t1; : : : ; tn} and their associated categories (interchangeably called dimensions) Du = {d1; : : : ;dn} (for example, grocery, dining, etc.) [(that is, “Tu” and “Du” are vector representations of cardholders and merchants based on relationships learned from transaction history)]”);
extract groups of representations from the representation space by grouping together representations corresponding to a same merchant category code (MCC) (Manzoor, Fig. 4(c), teaches purchase categories and MCCs [Examiner annotations in dashed-line text boxes]:
PNG
media_image1.png
675
621
media_image1.png
Greyscale
Manzoor, left column of p. 1925, “2. Problem Overview,” first partial paragraph, teaches “Transaction categories are derived from the Merchant Category Codes associated with transactions and form a finite set D [(that is, extract groups of representations from the representation space by grouping together representations corresponding to a same merchant category code (MCC))]”);
train . . . [a model] to learn transaction patterns of accounts by feeding, in temporal order (Manzoor, left column of p. 1926, “4. Modeling Purchase Behavior,” first paragraph, teaches “[f]ormally, a temporal point process is a stochastic process, the realization of which is an ordered sequence of event timestamps {ti} ⊂ [0, ∞) [(that is, an “ordered sequence” is feeding, in temporal order)]”), transactions associated with the merchants and the cardholders, the groups of representations, and contexts describing loyalty-related conditions at times of the transactions to the [model] (Manzoor, left column of p. 1925, “3. Data,” first & second paragraph, teaches “[o]ur partner bank provides a variety of financial services to customers across the United States, such as checking, savings, loan, credit and debit accounts. In this work, we focus on customers who hold prepaid card accounts. . . . Since these individuals are less likely to have other accounts (unlike others who could potentially have a number of credit cards across different banks), data collected from prepaid card customers represent a near-complete picture of their financial activities and is thus very suitable for modeling spending behaviors [(that is, transactions associated with the merchants and the cardholders, the groups of representations, and contexts describing loyalty-related conditions)]”),
wherein the transaction patterns correspond to different MCCs and transaction amount bins (Manzoor, left column of p. 1925, “3. Data,” second paragraph, teaches “[e]ach transaction is associated with a dollar amount [(that is, transaction amount bins)], a timestamp (at the granularity of seconds) and a Merchant Category Code (MCC) [(that is, wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins)]”), . . . ;
* * *
train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins (Manzoor, left column of p. 1925, “2. Problem Overview,” first full paragraph & second paragraph, teaches “[w]e now define the Coupon problem as that of one-step-ahead, next purchase prediction. Given a coupon forecasted by any method solving Coupon, for the probability of the coupon being redeemed to be high, each of the following intermediate results must be accurate: (1) The probability densities of the predicted purchase times
t
^
n
+
1
must be high at the actual purchase times tn+1. (2) Point predictions of the predicted purchase time must minimize the absolute prediction error
t
^
n
+
1
-
t
n
+
1
. (3) The predicted coupon interval
t
^
n
+
1
s
t
a
r
t
,
t
^
n
+
1
e
n
d
must trap the actual purchase time, i.e.,
t
n
+
1
∈
t
^
n
+
1
s
t
a
r
t
,
t
^
n
+
1
e
n
d
. (4) The probability density of the predicted purchase category
d
^
n
+
1
must be high at the actual purchase category dn+1”; Manzoor, right column of p. 1927, “4.4 Learning and Prediction,” first paragraph, teaches “[c]onsider a sequence of timestamps {t1; : : : ; tn} in an observation window [0;T) where the dimension of each timestamp ti is denoted by d(ti) ∈ {1; : : : ;D}. The loglikelihood of this sequence with respect to any temporal point process can be written directly in terms of its conditional intensity, . . . .);
generate predictions including future times of future transactions involving a user account, and specific merchants expected to coincide with the future times based on the probability density function (Manzoor, Fig. 3, teaches constructs an interval around a predicted next purchase time and over merchant categories [Examiner annotations in dashed-line text boxes]:
PNG
media_image2.png
270
1126
media_image2.png
Greyscale
Manzoor, Fig. 3 caption, teaches that “(top) Timeline for user u with ordered transactions
t
i
u
,
d
^
i
u
i
=
1
n
, where
t
i
u
are timestamps and
d
i
u
are merchant categories, and (bottom) an example realization. Our model RUSH! constructs an interval (highlighted in yellow) around the predicted next purchase time (
t
^
n
+
1
u
, black cross) [(that is, generate predictions including future times of future transactions involving a user account, as well as a predictive distribution over merchant categories [(that is, an specific merchants expected to coincide with the future times based on the probability density function)], based on which it delivers a personalized, time-limited digital coupon to user u”) . . . ;
filter irrelevant information including spam from the user based on the predictions of the probability density function (Manzoor, left column at p. 1929, “5. Model Evaluation,” last partial paragraph, teaches “[w]e also consider the impact of only delivering coupons for which we are confident about redemption. Intuitively, delivering only high-confidence coupons conservatively avoids the costs of spamming consumers [(that is, “only high-confidence coupons” is to filter irrelevant information including spam from the user based on the predictions of the probability density function)], but trades off against potentially larger gain from serving more coupons”);
automatically transmit a benefit to an owner of the user account based on the predicted future times and the specific merchants (Manzoor at p. 1924, “1. Introduction,” first partial paragraph, teaches “[o]ur key objective hence is to maximize the redemption rate of delivered coupons via data-driven personalization and targeting. We focus specifically on time-limited coupons characterized by a delivery time, duration, and a merchant category [(that is, automatically transmit a benefit to an owner of the user account based on the predicted future times and the specific merchants)]”).
Though Manzoor teaches the model parameters expressed in an excitation matrix, Manzoor, however, does not explicitly teach that the model is a LSTM RNN. Also, Manzoor does not explicitly teach –
* * *
[train . . . a model, wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,] the contexts include vectors representing encoded parameters including
parameter a corresponding to a category class,
parameter b corresponding to a class of progress percentage,
parameter c corresponding to a class of proximity, and
category d corresponding to a class of historical redemption;
* * *
But Yan teaches a “touchpoint attribution attention neural network includes an embedding layer, a recurrent neural network (RNN)/long short-term memory (LSTM) layer [(that is, a LSTM RNN)],” (Yan ¶ 0026), and also teaches -
* * *
[train . . . a model], wherein the transaction patterns correspond to different merchant category codes (MCCs) and transaction amount bins,] the contexts include vectors representing encoded parameters (Yan ¶ 0097 teaches “encoded touchpoint vectors 404, shown as x1, x2, . . . xT in FIG. 4A [(that is, contexts include vectors representing encoded parameters )]”) including
parameter a corresponding to a category class (Yan ¶ 0035 teaches “the term product hereafter refers to both products and services and includes subscriptions, bundles, and on-demand/one-time purchasable products [(that is, parameter a corresponding to a category class)]”),
parameter b corresponding to a class of progress percentage (Yan ¶ 0080 teaches “’DI’ or display impression, ‘DC’ or display click” and “’FT’ or free trial sign-up, and ‘PS’ or paid search [(that is, parameter b corresponding to a class of progress percentage)]”),
parameter c corresponding to a class of proximity (Yan ¶ 0080 teaches “’ES’ or email sent, ‘EO’ or email opened, ‘EC’ or email clicked [(that is, ”parameter c corresponding to a class of proximity)]”), and
category d corresponding to a class of historical redemption (Yan ¶ 0080 teaches “conversion indicators (e.g., “C” or conversion and “NC” or non-conversion) [(that is, category d corresponding to a class of historical redemption)]”);
* * *
Manzoor and Yan are from the same or similar field of endeavor. Manzoor teaches the use of large-scale anonymized transaction records to model consumer spending and forecast future purchases, based on which we generate data-driven, personalized coupons. Yan teaches using a touchpoint attribution attention neural network to generate conversion predictions for target touchpoint sequences and to provide targeted digital content over specific digital media channels to client devices of individual users.
Thus, it would have been obvious to a person having ordinary skill in the art as of the effective filing date of the Applicant’s invention to modify Manzoor pertaining to personalized consumer incentives at a forecast future purchase with the context vectors of Yan.
The motivation to do so is because a “deep learning attribution system can generate and utilize a touchpoint attribution attention neural network to efficiently and accurately generate accurate touchpoint attributions for a digital content campaign as well as generate conversion predictions for future touchpoints in digital content campaigns. Moreover, by utilizing a trained touchpoint attribution attention neural network the deep learning attribution system can flexibly model interactions between different media channels, temporal effects, user characteristics, and control variables.” (Yan ¶ 0021).
Regarding claims 2, 10, and 18, the combination of Manzoor and Yan teach all of the limitations of claims 1, 9, and 17, respectively, as described above in detail.
Manzoor teaches -
wherein the processor is configured to execute the instructions to additionally learn the transaction pattern by:
learning a distribution of an expenditure pattern relating the account over time (Manzoor, Figs. 7(a)-7(c), teach spending patterns of an account over time [Examiner annotations in dashed-line text boxes]:
PNG
media_image3.png
478
554
media_image3.png
Greyscale
Manzoor, left column of p. 1927, “4.2 Incorporating Memory,” first partial & first full paragraph, teaches “Fig. 7(a) shows, given a purchase at time tn, the distribution of the inter-purchase time (tn+1 - tn) conditioned on the number of purchases in the 24 hours prior to tn. We observe that having more purchases in the previous 24 hours shifts the distribution towards smaller inter-purchase times [(that is, learning a distribution of an expenditure pattern relating the account over time)]. We also investigate the sequential correlation between purchase categories using the cross-correlation coefficient (CCF), which measures the similarity of two time series at different lags. Fig. 7 (b) and (c) show example CCF plots at various lags (-4 to 4 hours in 30 minute increments). We see from (b) that people often Commute 30 mins before and up to 2 hrs after Dining, and (c) suggests the correlation is symmetric between Retail vs. Grocery. From these case studies, we conclude that a strong sequential correlation exists between purchases in different categories”).
Regarding claims 3, 11, and 19, the combination of Manzoor and Yan teaches all of the limitations of claims 1, 9, and 17, respectively, as described above in detail.
Manzoor teaches -
wherein the processor is configured to execute the instructions to additionally train the LSTM RNN by:
providing information related to a cardholder of the account transacting at time t (Manzoor, Fig. 3, teaches providing a coupon offer to a cardholder at time t [Examiner annotations in dashed-line text boxes]:
PNG
media_image4.png
225
939
media_image4.png
Greyscale
Manzoor, Fig. 3 caption teaches “(top) Timeline for user u with ordered transactions
t
i
u
,
d
^
i
u
i
=
1
n
, where
t
i
u
are timestamps and
d
i
u
are merchant categories, and (bottom) an example realization. Our model RUSH! constructs an interval (highlighted in yellow) around the predicted next purchase time (
t
^
n
+
1
u
, black cross) [(that is, providing information related to a cardholder of the account transacting at time t)], as well as a predictive distribution over merchant categories, based on which it delivers a personalized, time-limited digital coupon to user u”);
providing information related to a merchant being transacted at by the cardholder of the user account at time t Manzoor, Fig. 3 caption & Fig. 3 above, teaches “(top) Timeline for user u with ordered transactions
t
i
u
,
d
^
i
u
i
=
1
n
, where
t
i
u
are timestamps and
d
i
u
are merchant categories, and (bottom) an example realization. Our model RUSH! constructs an interval (highlighted in yellow) around the predicted next purchase time (
t
^
n
+
1
u
, black cross), as well as a predictive distribution over merchant categories, based on which it delivers a personalized, time-limited digital coupon to user u [(that is, “the predictive distribution over merchant categories” is providing information related to a merchant being transacted at by the cardholder of the user account at the time t)]”); and
providing a context of the cardholder of the user account and the merchant at time t (Manzoor, left column of p. 1927, “4.2 Incorporating Memory,” first paragraph, teaches “the sequential correlation between purchase categories using the cross-correlation coefficient (CCF), which measures the similarity of two time series at different lags. Fig. 7(b) and (c) show example CCF plots at various lags (-4 to 4 hours in 30 minute increments). We see from (b) that people often Commute 30 mins before and up to 2 hrs after Dining, and (c) suggests the correlation is symmetric between Retail vs. Grocery. From these case studies, we conclude that a strong sequential correlation exists between purchases in different categories [(that is, the “CCF” is providing a context of the cardholder of the user account and the merchant at time t)]”).
Regarding claims 4, 12, and 20, the combination of Manzoor and Yan teaches all of the limitations of claims 1, 9, and 17, respectively, as described above in detail.
Manzoor teaches -
wherein the processor is configured to execute the instructions to additionally generate the predictions by:
predicting transactions behavior of the user account during a period (Manzoor, left column of p. 1927, “4.2 Incorporating Memory,” first paragraph, teaches “the sequential correlation between purchase categories using the cross-correlation coefficient (CCF), which measures the similarity of two time series at different lags. Fig. 7(b) and (c) show example CCF plots at various lags (-4 to 4 hours in 30 minute increments) [(that is, “-4 to 4 hours” is during a period)]. We see from (b) that people often Commute 30 mins before and up to 2 hrs after Dining, and (c) suggests the correlation is symmetric between Retail vs. Grocery. From these case studies, we conclude that a strong sequential correlation exists between purchases in different categories [(that is, the “CCF” is predicting transactions behavior of the user account during a period)]”).
Regarding claims 5 and 13, the combination of Manzoor and Yan teaches all of the limitations of claims 1 and 9, respectively, as described above in detail.
Yan teaches -
wherein the processor is configured to execute the instructions to additionally change a state of the user account by:
linking the user account to a vector attribute of the future transactions involving the user account (Yan ¶ 0071 teaches that “the touchpoint attributions 210 include a weight, coefficient, number, or other values indicating how influential each touchpoint in the touchpoint sequence 204 was in leading to the reported conversion. In many embodiments, the sum of touchpoint attributions 210 adds to one. For example, the deep learning attribution system 104 determines that the first touchpoint 202 a (i.e., display impression) has an attribution value of 15%, the second touchpoint 202 b (i.e., email) has an attribution value of 35%, and the free trial sign-up has an attribution scale of 50% [(that is, the “touchpoint attribute sum” is a vector attribute)]. In alternative embodiments, the sum of touchpoint attributions 210 does not add to one or is above one”; Yan ¶ 0076 teaches a correlated marker where “each touchpoint includes a touchpoint identifier, a user identifier, and an interaction time (e.g., timestamp) [(that is, linking the user account to a vector attribute of the future transactions involving the user account for the correlated marker )]”).
Regarding claims 6 and 14, the combination of Manzoor and Yan teaches all of the limitations of claims 5 and 13, respectively, as described above in detail.
Manzoor teaches –
wherein the benefit includes at least one of a discount, offer, conditional reward, or incentive of a merchant (Manzoor, left column of p. 1924, “1. Introduction,” first partial paragraph, teaches to “maximize the redemption rate of delivered coupons via data-driven personalization and targeting. We focus specifically on time-limited coupons characterized by a delivery time, duration, and a merchant category. Time constraints are widely believed to accelerate coupon redemption by invoking feelings of urgency and scarcity [1, 16]. [(that is, wherein the feature benefit includes at least one of a discount, offer, conditional reward, or incentive of a merchant)]”).
Regarding claims 8 and 16, the combination of Manzoor and Yan teach all of the limitations of claims 1 and 9, respectively, as described above in detail.
Manzoor teaches –
wherein the benefit includes an existing offer or a future offer provided by a merchant (Manzoor, left column of p. 1924, “(1) Problem Formulation: Promoting Digital Coupons,” first paragraph, teaches “we formalize the problem of promoting personalized time-limited coupons to bank customers, which puts to use a large collection of transactions data from a national bank. Done effectively, such an application is expected to benefit all parties; by helping the customers save, the bank to raise profit from commissions, and the participating merchants to receive foot traffic”) associated with the MCCs (Manzoor, right column of p. 1925, “3. Data, Preprocessing,” first paragraph, teaches that “we start with the 2-digit grouping and manually split or combine categories as appropriate to construct an MCC-category mapping that comprises 10 broad purchase categories (listed in Fig. 4 (c)) [(that is, associated with the MCCs)]”)
Response to Arguments
9. Examiner has fully considered the Applicant’s arguments and responds below accordingly.
Section 101
10. Under Section 101, “the Office asserts that claim 1 recites ‘observations, evaluations, judgments, and opinions’ (Office Action, page 2). Applicant traverses this assertion. Representative claim 1 recites:
A system for managing a payment network, comprising:
a memory configured to store instructions of a predictor model; and
a transaction manager comprising a processor configured to execute the instructions to:
generate, using a representation learner, a representation space comprising vector representations of cardholders and merchants based on relationships learned from transaction history;
extract groups of representations from the representation space by grouping together representations corresponding to a same merchant category code (MCC);
train a long short-term memory recurrent neural network (LSTM RNN) to learn transaction patterns of accounts by feeding, in temporal order, transactions associated with merchants and cardholders, the groups of representations, and contexts describing loyalty-related conditions at times of the transactions to the LSTM RNN, wherein the transaction patterns correspond to different MCCs and transaction amount bins, the contexts include vectors representing encoded parameters including parameter a corresponding to a category class, parameter b corresponding to a class of progress percentage, parameter c corresponding to a class of proximity, and category d corresponding to a class of historical redemption;
train a probability density function of the LSTM RNN based on the vectors and the contexts, wherein the probability density function models times of future transactions and MCC bins;
generate predictions including future times of future transactions involving a user account, and specific merchants expected to coincide with the future times based on the probability density function of the LSTM RNN;
filter irrelevant information including spam from the user based on the predictions of the probability density function; and
automatically transmit a benefit to an owner of the user account based on the predicted future times and the specific merchants.
(Emphasis added [by Applicant]).” (Response at pp. 9-10).
Examiner Response:
Examiner respectfully submits that the claims are directed to an abstract idea, as set out above in detail. Under Step 2A Prong One, the rejection identifies the judicial exception (for example, an abstract idea) by referring to what is recited (that is, set forth or described) in the claim and explain why it is considered an exception. Specifically, where the claim is directed to an abstract idea, the rejection identifies the abstract idea as it is recited (i.e., set forth or described) in the claim and explains why it is an abstract idea. (see MPEP § 2106.07(a)).
For example, relating to exemplar claim 1, the activities of “generate a representation space, “extract groups of representations,” “generate predictions,” and “filter irrelevant information,” include limitations that can practically be performed in the human mind, including, for example, observations, evaluations, judgments, and opinions, and accordingly, are mental processes, (MPEP § 2106.04(a)(2) sub III), which is one of the groupings of abstract ideas. (MPEP § 2106.04(a)(2)).
Thus, the claims recite an abstract idea, as set out above in detail.
11. Under Step 2A Prong Two, Applicant submits that “Even if the Office maintains that claim 1 is drawn towards a judicial exception, Applicant submits that claim 1 should be found to be patent eligible at least at Prong Two by reciting additional elements to integrate the exception into a practical application.
Claim 1 is amended to emphasize features relating to training a LSTM RNN, which specifically is a machine learning model (a specific computer technology) as noted above in the emphasized features. Moreover, claim 1 further includes features relating to the probability density function of the LSTM RNN, and reduce network traffic (e.g., ‘filter irrelevant information . . .’).
The specification states
‘[l]oyalty programs that provide benefits to cardholders increases trust in credit card companies and other financial entities. However, loyalty offers and other incentives are often extended to customers who have demonstrated no interest in the good or services related to those offers. This provides cardholders with a sub-optimal user experience and diminishes the scope of increasing customer engagement’
(¶ [0002]). Further, the specification states ‘[s]pam offers serve as a constant source of inconvenience to account holders and reduce productivity.’ (¶ [0039]).
Examples herein solve the above problems and include a technical solution to solve the above. . . . This is a specific improvement to the functioning of a computer-implemented prediction model. The LSTM RNN and probability density function operate together to generate conditional probability distributions of future transaction times and specific merchants expected to coincide with the future times (¶¶ [0038]-[0039]), which conventional systems cannot do. Moreover, examples reduce network traffic.” (Response at pp. 9-11 (referring to Specification ¶¶ 0062, 0076, 0080 & 0039)).
Regarding Desjardins, Applicant submits that “Claim 1 of the present application is similar to those of Ex Parte Desjardins. For example, one improvement identified in the Specification of Ex Parte Desjardins is to ‘effectively learn new tasks in succession whilst protecting knowledge about previous tasks.’ When evaluating the claim as a whole, the Office discerned at least the following limitation of independent claim 1 that reflects the improvement: ‘adjust the first values of the plurality of parameters to optimize performance of the machine learning model on the second machine learning task while protecting performance of the machine learning model on the first machine learning task.’ The Office was persuaded that constitutes an improvement to how the machine learning model itself operates, and not, for example, the identified mathematical calculation.
Claim 1 similarly recites several features relating to training, to improve the overall efficiency, training process, and predictions. Doing so results in a technical enhancement to LSTM RNN and a probability density function. As such, the claims are integrated into a practical application.” (Response at pp. 11-12).
Examiner Response:
Examiner respectfully disagrees because under Step 2A Prong Two, the rejection identifies any additional elements (specifically point to claim features/limitations/steps) recited in the claim beyond the identified judicial exception; and evaluate the integration of the judicial exception into a practical application by explaining that the claim as a whole, looking at the additional elements individually and in combination, does not integrate the judicial exception into a practical application using the considerations set forth in MPEP §§ 2106.04(d), 2106.05(a)- (c) and (e)- (h). (MPEP § 2106.07(a)).
Under Step 2A Prong Two, “integration” may be based on the improvements in the functioning of a computer or an improvement to any other technology or technical field. (MPEP § 2106.04(d)(1)). The evaluation requires, [i]n sum, that (1) the specification should be evaluated to determine if the disclosure provides sufficient details such that one of ordinary skill in the art would recognize the claimed invention as providing an improvement. [hereinafter “Leg 1”]. Next, (2) if the specification sets forth such an improvement, the claim must be evaluated to ensure that the claim itself reflects the disclosed improvement. [hereinafter “Leg 2”].
Under Leg 1, Applicant points out that “Examples herein solve the above problems and include a technical solution to solve the above. . . . This is a specific improvement to the functioning of a computer-implemented prediction model. The LSTM RNN and probability density function operate together to generate conditional probability distributions of future transaction times and specific merchants expected to coincide with the future times (¶¶ [0038]-[0039]), which conventional systems cannot do. Moreover, examples reduce network traffic.” (Response at pp. 9-11 (referring to Specification ¶¶ 0062, 0076, 0080 & 0039) (emphasis added by Examiner)).
With regard to a “probability density function” and a “LSTM RNN,” the Specification recites
the RNN model be a long short-term memory (LSTM) recurrent neural network. An LSTM RNN is a type of deep-learning technique that uses a trained probability density function to predict, in this case, future transactions that are likely to occur at merchants and their corresponding transaction amount bin and specific accounts in the network. Based on these predictions, actions can be taken preemptively to target account holders with incentives within the payment network they otherwise would not have had or known about.
(Specification ¶ 0039 (emphasis added by Examiner)). With the activity of “filtering,” the Specification continues, in which the
The predictions may also filter out potentially irrelevant information (e.g., spam), e.g., incentives relating to merchants which an account holder has demonstrated no interest in in the past based on transactional and redemption history. Spam offers serve as a constant source of inconvenience to account holders and reduce productivity
(Specification ¶ 0039 (emphasis added by Examiner)). That is, an intended result of a prediction is to “filter out potentially irrelevant information.”
The Specification recites, in reference to Figure 2, that “the datasets 210 may be input to train the probability density function of the neural network on an account-by-account basis, or datasets for multiple accounts . . . may be input to train the density function.” (Specification ¶ 0042). Also, the Specification recites training of the probability density function through a temporal point process (TPP) model:
Training of the probability density function (learned by the TPP model) includes feeding three important types information at each time step (t) to the model. The first type of information includes a cardholder transacting at time t. The second type of information includes a corresponding merchant being transacted at by that cardholder at time t. The third type of information includes a context of the involved cardholder and merchant at time t, which is generated using the redemption history database. This information is fed to the TPP model in a chronological order to learn the probability density function.
(Specification ¶ 0055).
In other words, the Specification supports that the LSTM-RNN and probability density function are recited at a high-level of generality, and accordingly are generic computer components. Training the models (LSTM-RNN, probability density function), without more, is the use of a generic computer components to implement the abstract idea. (MPEP § 2016.05(f)), that does not amount to significantly more than the abstract idea under Step 2A Prong Two, nor are significantly more than the abstract idea under Step 2B. (see, e.g., claim 1, lines 9-17 (“training a LSTM RNN), claim 1, lines 18-20 (“training a probability density function of the LSTM RNN”)).
In contrast, under Desjardins, the Appeals Review Panel determined that the specification identified improvements as to how the machine learning model itself operates, including training a machine learning model to learn new tasks while protecting knowledge about previous tasks to overcome the problem of “catastrophic forgetting” encountered in continual learning systems. (Advance notice of change to the MPEP in light of Ex Parte Desjardins at p. 2 (05 December 2025) [hereinafter Advance Notice] (emphasis added by Examiner)).
Applicant’s disclosure does recite, for example, that “[t]he predictions may be used as a basis of increasing efficiency and productivity of the network, as well as allow the network to better serve its participating entities in transacting business,” (Specification ¶ 0029), and “the ability to target incentives to relevant account holders inures both to the benefit of the account holders and merchants, and also to the credit card company through an increase in loyalty of its card users.” (Specification ¶ 0039).
Though an abstract idea may be improved, it remains an abstract idea. Also, the additional elements of the claims are generic computer components (e.g., processor, memory, LSTM-RNN, probability density function) that are used to implement the abstract idea, (MPEP § 2106.05(f)), and also additional elements of post-processing communications relating to the abstract idea of “predicting.” Accordingly, the claims are subject-matter ineligible, as set out above in detail.
12. Under Step 2B, Applicant submits “Even assuming, solely for the sake of argument, that the Examiner disagrees with the above, the claims still satisfy Step 2B because the claim includes technical elements that amount to ‘significantly more’ than any alleged abstract idea. As claimed, several features are recited in the claims:
1) A representation learner;
2) A LSTM RNN; and
3) A probability density function.
The three features noted above operate together to reduce irrelevant information (e.g., spam) thereby reducing internet traffic. The claims are similar to Bascom stated that adding a specific limitation other than what is well-understood, routine, conventional activity in the field, or adding unconventional steps that confine the claim to a particular useful application, e.g., a nonconventional and non-generic arrangement of various computer components for filtering Internet content qualified as ‘significantly more’ (BASCOM Global Internet v. AT&T Mobility LLC, 827 F.3d 1341, 1350-51, 119 USPQ2d 1236, 1243 (Fed. Cir. 2016)) (MPEP § 2106.05(d)).’” (Response at p. 12).
Examiner Response:
Examiner respectfully disagrees. Under Step 2B, the rejections explain why the additional elements, taken individually and in combination, do not result in the claim, as a whole, amount to significantly more than the identified judicial exception. Further, for certain elements reciting well-understood, routine, and conventional activities in the relevant field, the rejection supports such conclusion in writing with one of the four options specified in Subsection III. (see MPEP § 2106.07(a)) & MPEP § 2106.07(a) sub III).
Generally, though an abstract idea may be improved in terms of efficiency or speed of execution, it remains an abstract idea. The additional elements of the claims are generic computer components (e.g., processor, memory, LSTM-RNN, probability density function) that are used to implement the abstract idea, (MPEP § 2106.05(f)), and also, the additional elements of post-processing communications relating to the abstract idea of “predicting.” Accordingly, the claims are subject-matter ineligible, as set out above in detail.
Section 103
13. Applicant submits “Yan does not appear to disclose or suggest a transaction as claimed in combination with the other claimed features. That is, the cited art does not appear to disclose or suggest ‘generate predictions including future times of future transactions involving a user account, and specific merchants expected to coincide with the future times based on the probability density function of the LSTM RNN.’” (Response at pp. 14-15).
Examiner Response:
Examiner agrees that Yan does not teach the limitations pointed to by Applicant. In this regard, Examiner relies upon the teachings of Manzoor, as set out above in detail.
Conclusion
14. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
(Dutta et al., “HawkesEye: Detecting Fake Retweeters Using Hawkes Process and Topic Modeling,” IEEE (2020)) teaches a novel classifier, called HawkesEye which makes predictions based on a temporal window, in contrast to existing approaches which require a graph-like relationship between tweet entities, or the presence of the entire retweeting timeline of a retweeter. HawkesEye utilizes both temporal and textual information using a class-specific topic model and Hawkes processes.
15. Any inquiry concerning this communication or earlier communications from the Examiner should be directed to KEVIN L. SMITH whose telephone number is (571) 272-5964. Normally, the Examiner is available on Monday-Thursday 0730-1730.
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, KAKALI CHAKI can be reached on 571-272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/K.L.S./
Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights