DETAILED ACTION
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
Applicant’s communications filed on 1/28/2026 have been considered.
Claims 1, 4-5, 7, 9-11, 14-15, 17, 19-21, and 23 have been amended.
Claims 3 and 13 have been canceled.
Claims 25 and 26 are newly added.
Claims 1, 4-7, 9-11, 14-17, and 19-26 are currently pending and have been examined.
Indication of Subject Matter Overcoming Prior Art
Claim 1 would be allowable if rewritten or amended to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office Action. Claims 4-10 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office action and to include all of the limitations of the base claim and any intervening claims. Claim 16 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office action and to include all of the limitations of respective base claim 11 and any intervening claims. Claim 22 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office action and to include all of the limitations of respective base claim 21 and any intervening claims.
Response to Arguments
Applicant’s arguments filed with respect to the rejection of claims under 35 USC 101 have been fully considered but they are not persuasive.
Applicant argues on pages 12-13 that the previous Office Action (filed 11/5/2025) fails to properly respond to the 101 arguments presented in the previous response. Applicant further argues that the previous Office Action fails to explain why the claims do not represent a technical improvement to conventional machine learning approaches thar require all factors to be accounted by machine learning. This argument has been considered but is not persuasive. A response to the cited argument can be found on pages 16-17 of the previous Office Action, where it was discussed that the additional elements of the claims, including building a machine learning model in real-time to automatically determine rankings, and wherein the machine learning model is built in real-time further based on: identifying a first sensitive Weighted Alternate Least Squares (WALS) model, identifying a second sensitive WALS model, and identifying a combination of the first sensitive WALS model and the second sensitive WALS model, did not represent any technical improvement to a computer or other technology, encompassing conventional machine learning approaches, but rather represented an improvement to the commercial task of providing ranked recommendations, applied utilizing a generic computing environment. This response encompasses the argument of whether the claims represent a technical improvement to conventional machine learning approaches, as conventional machine learning approaches fall within the realm of a computer or other technology. Nevertheless, an additional response to Applicant’s arguments with respect to Step 2A Prong Two, are further discussed below. It is noted that this action is mailed as a Non-Final Action.
Applicant argues on pages 14-15 that the claims do not recite an abstract idea because amended claim 1 is similar to USPTO’s example 39, in that the claim does not recite any judicial exception. This argument has been considered but is not persuasive. With regards to Example 39, as discussed in the previous Office Action (see page 15), it was determined the claims do not recite a judicial exception, as they are directed to a computer-implemented method of training a neural network for facial detection. With regards to the instant claims, Examiner maintains that the claims are not similar to those in Example 39, as they have been identified as reciting limitations falling under “Certain Methods of Organizing Human Activity”, as they recite commercial/legal interactions including advertising, marketing or sales activity, or business relations, including displaying ranked item information, as discussed below. Accordingly, the instant claims recite a judicial exception, unlike those set forth in Example 39.
With regards to Applicant’s argument on page 15 that the claims do not mention recommendations, this argument has been considered but is not persuasive. The claims currently stand rejected under 101 for being directed to the “Certain Methods of Organizing Human Activity” grouping of abstract ideas, as noted above, for reciting commercial/legal interactions, including displaying ranked item information (i.e., a recommendation). This is further evidenced by the specification disclosing that the ranking of items for display as a ranking of new items for the user to add to the cart for the current user session (see at least [0053]).
Accordingly, Examiner maintains that the claims recite an abstract idea.
Applicant further argues on pages 15-17 that the claims integrate the abstract idea into a practical application. Applicant argues that the claims provide improvements including “significantly reducing processing resources required and improving machine learning performance relative to conventional machine learning approaches that involve maintaining multiple machine learning models before each session, as well as conventional machine learning approaches that require a machine learning model to be re-executed to create the same type of output previously created by the machine learning model”. This argument has been considered and is not persuasive.
If it is asserted that the invention improves upon conventional function of a computer, or upon conventional technology or technological processes, a technical explanation as to how to implement the invention should be present in the specification. That is, the disclosure must provide sufficient details such that one of ordinary skill in the art would recognize the claimed invention as providing an improvement. Although the specification need not explicitly set forth the improvement, it must describe the invention such that the improvement would be apparent to one of ordinary skill in the art. Conversely, if the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), the examiner should not determine the claim improves technology (see MPEP 2106.05(a); MPEP 2106.04(d)(1)).
Applicant’s specification does not provide the requisite detail necessary such that one of ordinary skill in the art could recognize the claimed invention as providing an improvement. With respect to Applicant’s argument that the claims provide an improvement over conventional machine learning approaches, Applicant’s specification does not provide sufficient detail with respect to building a machine learning model for the current session in real-time, such that one of ordinary skill in the art would understand that the claims provide a technical improvement over conventional machine learning approaches, and is specific only in their use in facilitating the abstract idea of displaying ranked item information. Nor does the specification provide technical detail of how building a machine learning model in real-time provides an improvement to the functioning of the claimed machine learning models themselves, differently than from generic machine learning processing (see at least specification [0052-0056][0060][0063][0068-0070]). For example, these paragraphs discuss machine learning processing at a high level, without providing the technical detail required such that one of ordinary skill in the art would recognize an improvement to technology. Rather, these paragraphs amount to a bare assertion of an improvement sans sufficient detail to demonstrate that Applicant has provided the alleged improvement to the technical field.
With regards to Applicant’s argument that the claims provide significantly reduced processing resources required, Applicant’s specification similarly does not provide sufficient technical detail with respect to reducing processor use, such that one of ordinary skill in the art would understand that the claims provide a technical improvement to computer processing (see at least [0068-0070]), but rather discusses the machine learning models at a high level without providing the technical detail required such that one of ordinary skill in the art would recognize a technical improvement.
It is further noted that creating a reduced set of items falls under the abstract idea, and therefore recites “Certain Methods of Organizing Human Activity”, as discussed below, and accordingly does not represent an improvement to machine learning. Furthermore, creating a reduced set of information for machine learning processing does not represent an improvement to the functioning of the machine learning model, or other technical area, as it merely describes processing/filtering information prior to use of said model.
The alleged improvements by Applicant are at best bare assertions of an improvement sans sufficient detail to demonstrate that Applicant has provided the alleged improvement to machine learning, computer processing, or another technical area. Accordingly, the claims to not integrate the abstract idea into a practical application, and the rejection has been maintained.
With regards to Applicant’s argument on page 17 that new dependent claims 25 and 26 are patent eligible for the reasons discussed with respect to claim 1, this argument has been considered but is not persuasive, as the 101 rejection of claim 1 has been maintained for the reasons discussed above. Accordingly, new claims 25 and 26 stand rejected under 101 for the reasons discussed below.
Indication of Subject Matter Overcoming Prior Art
Note: Claims 11, 14-15, 17, 19-21, and 23-26 have been rejected under 35 U.S.C. 103, in light of the amendments to claims 11 and 21 broadening the scope of the claims, as discussed below. The Indication of Subject Matter Overcoming Prior art pertains only to claims 1, 4-10, 16 and 22.
Claim 1 would be allowable if rewritten or amended to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office Action. Claims 4-10 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office action and to include all of the limitations of the base claim and any intervening claims. Claim 16 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office action and to include all of the limitations of respective base claim 11 and any intervening claims. Claim 22 would be allowable if rewritten to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office action and to include all of the limitations of respective base claim 21 and any intervening claims. The following is a statement of reasons for the indication of allowable subject matter:
Upon review of the evidence at hand, it is concluded that the totality of the evidence in combination, neither anticipates, reasonably teaches, nor renders obvious the below noted features of the Applicant’s invention as a whole, as the noted features amount to more than a predictable use of elements in the prior art. The allowable features are as follows:
building a machine learning model for the current session in real-time based on… a threshold associated with an aggregate value of the items to automatically determine new items to be selected, via the GUI of the user device, as additions to the electronic representation to satisfy the threshold associated with the aggregate value of the items;
wherein the machine learning model for the current user session in built in real-time further based on:
identifying a first sensitive Weighted Alternate Least Squares (WALS) model for a cart context of routine items;
identifying a second sensitive WALS model for a cart context of non-routine items; and
identifying a combination of the first sensitive WALS model and the second sensitive WALS model based on a cart context that includes routine items and non-routine items;
creating, based on creating the set of the new items using the machine learning model, a reduced set of the new items for the current session based on a respective time window score for each of the new items;
building a first sensitive Weighted Alternate Least Squares (WALS) model; and
building a second sensitive WALS model.
The most appropriate prior art of record includes Arora et al. (US 20210241343 A1), hereinafter Arora, Archak et al. (US 20210192596 A1), hereinafter Archak, Veettil et al. (US 20210295364 A1), hereinafter Veettil, Trepca et al. (US 20160140519 A1), hereinafter Trepca, Su et al. (US 20220277375 A1 A1), hereinafter Su, and NPL Reference U (Takacs et al., “Alternating Least Squares for Personalized Ranking”), hereinafter NPL Reference U.
Arora teaches a personalized recommendation system that can provide item recommendations to a user based on items that the user has included in a basket of selected items (Arora: [0040-0042]). Arora discloses receiving, from a database, data regarding one or more interactions with a graphical user interface (GUI) of a user device (Arora: [0037][0046-0049][0096][0123]). Arora further discloses analyzing selections, via the GUI of the user device, of items in an electronic representation for a current session to determine an online journey via one or more of a mobile application or a website, using the electronic representation (Arora: [Fig. 5][0027][0032][0073-0074]). Arora further discloses using a machine learning model for the current session in real-time based on the data regarding the one or more interactions with the GUI of the user device, the online journey via one or more of the mobile application or the website, a type corresponding to one or more types of the items, and additional information to automatically determine rankings of new items to be selected, via the GUI, as additions to the electronic representation (Arora: [0042][0051-0053][0109]). Arora further discloses creating a set of the new items using the machine learning model (Arora: [0046][0051-0053]), and creating, based on creating the set of the new items using the machine learning model, a reduced set of the new items for the current session based on a respective score for each of the new items and a respective status for each of the new items corresponding to whether each of the new items is selectable, via the GUI of the user device, to be added to the electronic representation (Arora: [0067-0068][0090][0094]). Yet Arora does not disclose determining that an online journey is one of: routine, corresponding to first items of the items that are routinely selected, non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to first items of the items that are routinely selected and the second items of items that are not routinely selected; building a machine learning model based on a threshold associated with an aggregate value of the items; wherein the machine learning model is built further based on: identifying a first sensitive Weighted Alternate Least Squares (WALS) model for a cart context of routine items; identifying a second sensitive WALS model for a cart context of non-routine items; and identifying a combination of the first sensitive WALS model and the second sensitive WALS model based on a cart context that includes routine items and non-routine items; creating a reduced set of the new items based on a respective time window score for each of the new items; building a first sensitive Weighted Alternate Least Squares (WALS) model, and building a second sensitive WALS model.
Archak teaches an online concierge system capable of determining key ingredients from items in a customer’s online shopping cart by mapping the items to generic items and removing non-ingredient items and staple items (Archak: [abstract]). Archak further discloses determining if a shopping journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected (Archak: [0032-0035]). Archak further discloses a cart context of routine items (Archak: [0034-0035]), a cart context of non-routine items (Archak: [0033]), and a cart context that includes routine and non-routine items (Archak: [0032-0035]). Yet Archak does not disclose the limitations regarding building a machine learning model based on a threshold associated with an aggregate value of the items; wherein the machine learning model is built further based on: identifying a first sensitive WALS model; identifying a second WALS model; identifying a combination of the first WALS model and the second WALS model; creating a reduced set of items based on a time window score, and building a first and second WALS model.
Veettil teaches a system and method for populating a basket associated with a first user in response to receiving a selection of a primary product at a computing device (Veettil: [0010]). Veettil further discloses building a machine learning model based on the aggregate value of the items in the electronic cart (Veettil: [0019][0052][0099]). Veettil further discloses Veettil further discloses determining rankings of new items to satisfy the threshold associated with the aggregate value of the items in the cart of the user (Veettil: [0019][0043]). Yet Veettil does not disclose the limitations regarding analyzing items in an electronic representation to determine that a journey is one of routine, non-routine, or mixed; identifying a first sensitive WALS model; identifying a second sensitive WALS model; identifying a combination of the first sensitive WALS model and the second sensitive WALS model; creating a reduced set of the new items based on a respective time window score for each of the new items; building a first sensitive Weighted Alternate Least Squares (WALS) model, and building a second sensitive WALS model.
Su teaches methods and systems for ranking recommended items based on user interactions within the same web session (Su: [abstract][0028]). Su further discloses reducing a size of a set of the new items before using the machine learning model (Su: [0029-0030][0049-0051]). Su further discloses managing a user’s shopping cart during a web session (Su: [0029]). Yet Su does not explicitly teach the limitations regarding determining that an online journey is one of routine, non-routine, or mixed; identifying a first sensitive WALS model and a second sensitive WALS model; identifying a combination of a first and second WALS model, creating a reduced set of items based on a time window score, or building a first and second sensitive WALS model.
NPL Reference U teaches using Alternating Least Squares modeling for personalized ranking (NPL Reference U: [Title][Page 1, Paragraphs 4-9][Page 3, Paragraphs 15-16]). NPL Reference U further discloses using a user-item interaction matrix factorization in combination with an alternating least squares (ALS) optimized in provide recommendations based on implicit feedback (NPL Reference U: [Page 3, Paragraphs 1-2]). U further discloses wherein machine learning models, such as ALS, include building the models (NPL Reference U: [Page 4, Paragraphs 15-16]). NPL Reference U further discloses using the ALS models to produce embeddings for both users and items (NPL Reference U: [Page 4, Paragraphs 15-19]). Yet NPL Reference U does not explicitly teach the limitations regarding a threshold associated with an aggregate value of the items to automatically determine new items to be selected, identifying a combination of a first and a second sensitive WALS model based on a cart context that includes routine items and non-routine items, and creating a reduced set of the new items based on a respective time window score for each of the new items.
While these references arguably teach the claimed limitations using a piecemeal analysis, these references would only be combined and deemed obvious based on knowledge gleaned from the applicant’s disclosure. Such a reconstruction is improper (i.e., hindsight reasoning). Accordingly, claim 1, taken as a whole, is indicated to be allowable over the cited prior art. The examiner emphasizes that it is the interrelationship of the limitations that renders these claims allowable over the prior art. Claims 4-10 depend from claim 1, respectively, and are therefore indicated as containing subject matter overcoming the prior art. Furthermore, claims 16 and 22 are indicated as containing subject matter overcoming the prior art, if rewritten to include the limitations of respective base claims 11 and 21, and any intervening claims.
Additionally, the Examiner further emphasizes the claims as a whole and herby asserts the totality of the evidence neither anticipates nor renders obvious the particular combination of elements as claimed. That is, the Examiner emphasized the claims as a whole and hereby asserts that the totality of evidence fails to set forth, either explicitly or implicitly, an appropriate rationale for combining or otherwise modifying the available prior art to arrive at the claimed invention. The combination of features as claimed would not be modifying the available prior art to arrive at the claimed invention. The combination of features as claimed would not be obvious to one of ordinary skill in the art because any combination of the evidence at hand to reach the combination of features as claimed would require a substantial reconstruction of Applicant’s claimed invention relying on improper hindsight bias.
It is hereby asserted by the Examiner that, in light of the above and in further deliberation over all of the evidence at hand, that claims 1, 4-10, 16 and 22 are allowable as the evidence at hand does not anticipate the claims and does not render obvious any further modification of the references to a person of ordinary skill in the art.
Claim Objections
Claim 11 is objected to because of the following informalities:
Regarding Claim 11, the claim recites “creating a ranking the new items by using the machine learning model”. It is noted that a typographical error was made with regards to this limitation, and for examination purposes, this limitation has been interpreted as “creating a ranking of the new items by using the machine learning model”. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
Claims 1, 4-7, 9-11, 14-17, and 19-26 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim 1 recites “determine that an online journey, via one or more of a mobile application or a website is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected”. The claim further recites “routine items” and “non-routine items”. The subject matter of the claim does not conform to the disclosure in such a manner in which one of ordinary skill in the art would recognize such method as being that which the Applicant adequately described as the invention or what the Applicant actually had possession of at the time of the filing. A review of the disclosure does not reveal the specific formula/algorithm used to determine items that are routinely selected from items that are not routinely selected. Utilizing “one or more machine learning models… to perform the method” is considered a black box formula/algorithm and is not sufficient to show possession of determining routine items, corresponding to items that are routinely selected, and non-routine items, corresponding to items that are not routinely selected (see Specification paragraphs [0050-0051][0055][0060][0068]). There must be some explanation of what the formula/algorithm is and how the model “determines” that an online journey is one of routine, non-routine, or mixed, based on analyzing selections, via the GUI of the user device, of items in an electronic representation. Disclosure of function alone is little more than a wish for possession and it does not satisfy the written description requirement. (See MPEP 2163: II(3)(a)(i), Eli Lilly, 119 F.3d at 1568, 43 USPQ2d at 1406). Moreover, a specification which does little more than outline goals applicant hopes the claimed invention achieves does not satisfy the written description requirement (see MPEP 2163: II(3)(a)(i), In re Wilder, 736 F.2d 1516, 1521, 222 USPQ 369, 372-73 (Fed. Cir. 1984)). It is noted that this is not an enablement rejection. Applicant’s failure to disclose any meaningful structure/algorithm as to how an online journey is one of: routine, non-routine, or mixed raises questions whether the Applicant truly had possession of this feature at the time of filing.
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1, 4-7, 9-11, 14-17, and 19-26 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites “determine that an online journey, via one or more of a mobile application or a website is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected”. The claim further recites “routine items” and “non-routine items”.
The metes and bounds of this claim is unclear inasmuch as one of ordinary skill in the art cannot determine how to avoid infringement of this claim because they are not apprised of what is to be determined as a “routine, corresponding to first items… that are routinely selected; non-routine, corresponding to second items… that are not routinely selected; or mixed, corresponding to the first items… that are routinely selected and the second items… that are not routinely selected”. When a term of degree is used in the claim, the examiner should determine whether the specification provides some standard for measuring that degree. See MPEP 2173.05(b). The specification does not disclose or provide some standard for measuring/determining whether items are routine or non-routine (see specification [0051][0055][0060][0068], which discusses that the method includes determining routine and non-routine items based on historical interaction information of users, but does not explicitly describe what makes those items considered routine and non-routine), rendering it unclear to how this is accomplished and what the metes and bounds of the claim are.
Dependent Claims 6, 15 and 22 similarly recite “routine items” and “non-routine items”, and are accordingly rejected for similar reasons.
Independent claims 11 and 19 recite substantially similar limitations to those discussed above with respect to claim 1, and accordingly inherit the deficiency noted above. Dependent claims 4-5, 7, 9-10, 14, 16-17, 19-20, and 23-26 inherit the deficiency noted in claims 1, 11 and 19.
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, 4-7, 9-11, 14-17, and 19-26 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claims recite an abstract idea. The judicial exception is not integrated into a practical application. The claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Under Step 1 of the Subject Matter Eligibility Test for Products and Processes, the claims must be directed to one of the four statutory categories. See MPEP 2106.03. Claims 1, 4-7, 9-10, and 25-26 are directed towards a machine. Claims 11, 14-17, and 19-20 directed towards a process. Claims 21-24 are directed towards a manufacture. Therefore, claims 1, 4-7, 9-11, 14-17, and 19-26 are directed to one of the four statutory categories (Step 1: YES, regarding claims 1, 4-7, 9-11, 14-17, and 19-26).
Under Step 2A of the MPEP, it is determined whether the claims are directed to a judicially recognized exception. See MPEP 2106.04. Step 2A is a two-prong inquiry.
Under Prong 1, it is determined whether the claim recites a judicial exception. In determining whether the claims are directed to a judicial exception, the claims are analyzed to evaluate whether the claims recite a judicial exception.
Taking Claim 1 as representative, claim 1 recites limitations that fall within the certain methods of organizing human activity groupings of abstract ideas, including:
receiving data regarding one or more interactions;
analyzing selections of items in a representation for a current session to determine that a journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected;
accessing a model for the current session in real-time based on the data regarding the one or more interactions, a type corresponding to one or more types of the items, and a threshold associated with an aggregate value of the items to determine new items to be selected as additions to the representation to satisfy the threshold associated with the aggregate value of the items;
wherein the model for the current user session is in real-time further based on:
identifying a cart context of routine items;
identifying a cart context of non-routine items; and
identifying a combination based on a cart context that includes routine items and non- routine items;
creating a set of the new items; and
creating, based on creating the set of the new items, a reduced set of the new items for the current session based on a respective time window score for each of the new items and a respective status for each of the new items corresponding to whether each of the new items is selectable to be added to the representation.
Claim 11 additionally recites:
receiving interaction information regarding one or more interactions;
analyzing items in a representation, for a current user session to determine that a journey, using the representation, is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected;
accessing a model for the current user session in real-time based on the interaction information, the journey, a type corresponding to one or more types of the items, and a threshold associated with an aggregate value of the items in the representation to determine rankings of new items to be selected, as additions to the representation to satisfy the threshold associated with the aggregate value of the items;
creating a ranking the new items; and
creating, based on creating the ranking of the new items, a re-ranking of the ranking of the new items based on a time window score and a status corresponding to an item, of the new items, being selectable to be added to the representation.
Claim 21 additionally recites the same abstract limitations as recited in claim 11, and additionally recites:
transmitting information based on the re-ranking of the new items.
Claims 1, 11 and 21, as exemplary, recites certain methods of organizing human activity, such as performing commercial interactions. See MPEP 2106.04(a)(2). The MPEP defines the “Certain Methods of Organizing Human Activity” grouping as including fundamental economic principles or practices (including hedging, insurance, mitigating risk); commercial or legal interactions (including agreements in the form of contracts; legal obligations; advertising, marketing or sales activities or behaviors; business relations); managing personal behavior or relationships or interactions between people (including social activities, teaching, and following rules or instructions) (see MPEP § 2106.04(a)(2). The abstract ideas recited in representative claim 1 are certain methods of organizing human activity because receiving data regarding interactions, analyzing selections of items in a representation for a current session to determine that a journey is one or: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected, accessing a model for the current session in real-time based on the interaction data, a type corresponding to one or more types of the items, and a threshold associated with an aggregate value of the items to determine new items to be selected as additions to the representation to satisfy the threshold associated with the aggregate value of the items, identifying a cart context of routine items, identifying a cart context of non-routine items, and identifying a combination based on a cart context that includes routine items and non-routine items, creating a set of the new items, and creating, based on creating the set of the new items, a reduced set of the new items for the current session based on a respective time window score and a respective status corresponding to whether each of the new items is selectable to be added to the representation is a commercial or legal interaction because it is an advertising, marketing or sales activity, or business relations.
Accordingly, under Prong One of Step 2A of the Alice/Mayo test, claims 1, 11 and 21 recite an abstract idea (Step 2A, Prong One: YES).
Under Step 2A (prong 2), if it is determined that the claims recite a judicial exception, it is then necessary to evaluate whether the claims recite additional elements that integrate the judicial exception into a practical application of that exception (see MPEP 2106.04). As stated in the MPEP, when “an additional element merely recites the words ‘apply it (or an equivalent) with the judicial exception, or merely uses a computer as a tool to perform an abstract idea,” the judicial exception has not been integrated into a practical application. In this case, representative claim 1 includes additional elements such as (additional elements are bolded):
A system comprising:
one or more processors; and
one or more non-transitory computer-readable media storing computing instructions that, when executed on the one or more processors, perform operations comprising:
receiving, from a database, data regarding one or more interactions with a graphical user interface (GUI) of a user device;
analyzing selections, via the GUI of the user device, of items in an electronic representation for a current session to determine that an online journey via one or more of a mobile application or a website is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected;
building a machine learning model for the current session in real-time based on the data regarding the one or more interactions with the GUI of the user device, the online journey via one or more of the mobile application or the website, a type corresponding to one or more types of the items, and a threshold associated with an aggregate value of the items to automatically determine new items to be selected, via the GUI of the user device, as additions to the electronic representation to satisfy the threshold associated with the aggregate value of the items;
wherein the machine learning model for the current user session is built in real-time further based on:
identifying a first sensitive Weighted Alternate Least Squares (WALS) model for a cart context of routine items;
identifying a second sensitive WALS model for a cart context of non-routine items; and
identifying a combination of the first sensitive WALS model and the second sensitive WALS model based on a cart context that includes routine items and non-routine items;
creating a set of the new items using the machine learning model; and
creating, based on creating the set of the new items using the machine learning model, a reduced set of the new items for the current session based on a respective time window score for each of the new items and a respective status for each of the new items corresponding to whether each of the new items is selectable, via the GUI of the user device, to be added to the electronic representation.
Claims 11 and 21 recite the same additional elements as recited in claim 1.
These additional elements are described at a high level in Applicant’s specification without any meaningful detail about their structure or configuration. As such, these computer-related limitations are not found to be sufficient to integrate the abstract idea into a practical application. Claims 1, 11 and 21 specifying that the abstract idea of displaying ranked item information is executed in a computer environment merely indicates a field of use in which to apply the abstract idea because this requirement merely limits the claims to the computer field, i.e., to execution on a generic computer. As such, under Prong Two of Step 2A of the Alice/Mayo test, when considered both individually and as a whole, the limitations of claims 1, 11 and 21 are not indicative of integration into a practical application (Step 2A, Prong Two: NO).
Since claims 1, 11 and 21 recite an abstract idea and fail to integrate the abstract idea into a practical application, claims 1, 11 and 21 are “directed to” an abstract idea (Step 2A: YES). Accordingly, the judicial exception is not integrated into a practical application.
Next, under Step 2B, examiners should evaluate additional elements individually and in combination to determine whether they provide an inventive concept (i.e., whether the additional elements amount to significantly more than the exception itself). In this case, the claims do not include additional elements that are sufficient to amount to significantly more than the judicial exception. Returning to representative claims 1, 11 and 21, taken individually or as a whole the additional elements of claims 1, 11 and 21 amount to no more than mere instructions to apply the exception using a generic computer and/or no more than a general link to a technological environment. For the same reason these elements are not sufficient to provide an inventive concept. Therefore when considering the additional elements alone, and in combination, there is no inventive concept in the claim, and thus the claim is not patent eligible (Step 2B: NO).
Dependent claims 4-7, 9-10, 14-17, 19-20, and 22-26, when analyzed as a whole, are held to be patent ineligible under 35 U.S.C. 101 because they do not add “significantly more” to the abstract idea. As for dependent claims 4-5, 7, 9, 14, 17, 19, 23 and 25, these claims recite limitations that further define the same abstract idea noted in independent claims 1, 11 and 21, and do not recite any additional elements other than what is disclosed in independent claims 1, 11 and 21. Therefore, claims 4-5, 7, 9, 14, 17, 19, 23 and 25 are considered patent ineligible for the reasons given above.
As for dependent claims 6, 10, 15-16, 20, 22, 24 and 26, these claims recite limitations that further define the abstract idea noted in independent claims 1, 11 and 21. Additionally, they recite the following additional limitations:
building the first sensitive WALS model to determine one or more of user embeddings or item embeddings for routine items; and
building the second sensitive WALS model to determine one or more of user embeddings or item embeddings for non-routine items;
including a first subset of the re-ranked ranking of the new items in a first portion of the GUI; and
including remaining ones of the re-ranked ranking of the new items in a second portion of the GUI;
building a sensitive Weighted Alternate Least Squares (WALS) model to determine one or more of user embeddings or item embeddings;
using, to improve operations of a serving layer, a backend server to build the machine learning model.
The additional elements of building the first sensitive WALS model; building the second sensitive WALS model; a first portion of the GUI; a second portion of the GUI; building a sensitive Weighted Alternate Least Squares (WALS) model; operations of a serving layer; and a backend server is recited at a high level of generality such that they amount to no more than instructions to apply the judicial exception in a generic technological environment. Even in combination, these additional elements do not integrate the abstract idea into a practical application and do not amount to significantly more than the abstract idea itself. Accordingly, under the Alice/Mayo test, claims 1, 4-7, 9-11, 14-17, and 19-26 are ineligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
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 11, 14, 19-21, 23-24, and 26 are rejected under 35 U.S.C. 103 as being unpatentable over previously cited Arora (US 20210241343 A1) in view of previously cited Archak (US 20210192596 A1), and further in view of previously cited Veettil (US 20210295364 A1).
Note: This action is made as a second non-final rejection, as a new grounds of rejection not previously made under 35 USC 103 have been deemed necessary, in light of the amendments to claims 11 and 21 broadening the scope of the claims.
Regarding Claim 11, Arora discloses A method, the method comprising ([0015-0016][0021]):
receiving interaction information regarding one or more interactions with a graphical user interface (GUI) of a user device ([Fig. 4]; [0046-0049] using past purchase data for users… over a one year time frame with user-item interactions; see [0096][0123] items selected for purchase via a user interface);
analyzing items in an electronic representation, for a current user session to determine an online journey, using the electronic representation ([Fig. 5]; [0073-0074] receiving a basket (virtual cart) for a user including items that have been selected by the user… each item in the basket can be grouped by L3 categories);
using a machine learning model for the current session in real-time based on the interaction information, the online journey, a type corresponding to one or more types of the items, and additional information in the electronic representation to automatically determine rankings of new items to be selected, via the GUI, as additions to the electronic representation ([0042] a within-basket recommender, which can suggest grocery items that go well with the items in a shopping basket (e.g., cart) of the user, such as milk with cereals, or pasta with pasta sauce… considering both (i) item-to-item compatibility within a shopping basket and (ii) user-to-item affinity; [0051-0053] determining a top k complementary items as outputs for an anchor item j and user in order to recommend one or more items that are personalized for a user; [0109]);
creating a ranking of the new items by using the machine learning model, as built ([0046] the triple embeddings model is trained to define cohesion scores; [0051-0053] the top k complementary items (i) can be determined by the triple embeddings model as outputs); and
creating, based on creating the ranking of the new items using the machine learning model, a re-ranking of the ranking of the new items based on a score and a status corresponding to an item, of the new items, being selectable to be added to the electronic representation ([0067-0068] the complementary category filtering technique can involve applying the lift scores to the complementary items, such that truly complementary items are boosted more, while popular yet unrelated items are boosted less, such that these latter items can drop lower in the score ranking and be effectively filtered out; [0094] cohesion scores determined for each recommended item are adjusted based on lift scores; see [0090]) (Note: items being identified as “complementary” and “popular”, for subsequent recommendation ranking using the cohesion sore, corresponds to re-ranking based on a status corresponding to an item being selectable to be added to the electronic representation).
Arora discloses analyzing items in an electronic representation, for a current user session to determine an online journey, using the electronic representation (Arora [Fig. 5][0073-0074]). However, Arora does not explicitly teach determining that an online journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected.
However, in the field of item recommendations (see at least Archak [abstract][0001-0002]), Archak, on the other hand, teaches determining that an online journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected ([Fig. 4]; [0032-0035] the cart analyzer receives items in a user’s online shopping cart… to determine key ingredients which are more commonly purchased… and to determine complementary ingredients that often cooccur in recipes with one or more of the key ingredients… to recommend to the customer).
The step of Archak is applicable to the method of Arora, as they share characteristics and capabilities, namely, they are directed to providing item recommendations according to machine learning analysis. It would have been obvious to one of ordinary skill in the art at the time of filing to modify the machine learning-based recommendation method as taught by Arora, to include determining that an online journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected, as taught by Archak. One of ordinary skill in the art at the time of filing would have been motivated to expand the recommendation method of Arora in order to determine recommended items based on items that have been added to an online shopping cart and their co-occurrence with other items, as well as give customers better suggestions of what to add to an order (Archak, [0001-0002]).
Arora further discloses using a machine learning model for the current session in real-time based on the interaction information, the online journey, a type corresponding to one or more types of the items, and additional information of the items in the electronic representation to automatically determine rankings of new items to be selected, via the GUI, as additions to the electronic representation (Arora [0042][0051-0053][0109]). Arora further discloses creating a re-ranking of the ranking of the new items based on a score ([0067-0068][0090][0094]). However, Arora in view of Archak does not explicitly teach building a machine learning model based on an aggregate value of the items to determine items to be selected to satisfy the threshold associated with the aggregate value of the items; and wherein a score is a time window score.
Additionally, in the field of item recommendations based on user selections (see at least Veettil [abstract][0010-0013]), Veettil, on the other hand, teaches building a machine learning model based on an aggregate value of the items to determine items to be selected to satisfy the threshold associated with the aggregate value of the items ([0052] generating a user-specific intent model to intake a set of basket characteristics (e.g., a price associated with the primary product)… as a user selects rewards offered to her over time, the computer system can record product characteristics (types, prices, promotional discounts, product groupings); see [0019] displaying purchase options to a user in response to a user’s selection of a product and the reward associated with the product (e.g., 20% discount)… the offer generation process is executed repeatedly until the basket size (e.g., the price of the basket) reaches a certain threshold value); and
wherein a score is a time window score ([0052] the intent model intakes a set of basket characteristics (e.g., a time of day) and outputs a predicted intent score for purchase of a particular secondary product).
The steps of Veettil are applicable to the method of Arora in view of Archak, as they share characteristics and capabilities, namely, they are directed to providing item recommendations based on user selections. It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak, to include building a machine learning model based on an aggregate value of the items to determine items to be selected to satisfy the threshold associated with the aggregate value of the items; and wherein a score is a time window score, as taught by Veettil. One of ordinary skill in the art at the time of filing would have been motivated to expand the recommendation method of Arora in view of Archak in order to determine recommended items based on items that have been added to an online shopping cart and their co-occurrence with other items, as well as give customers better suggestions of what to add to an order (Veettil, [0001-0002]).
Regarding Claim 14, Arora in view of Archak and Veettil teaches the limitations of claim 11.
Arora in view of Archak does not explicitly disclose determining a current total for the items; identifying an upper limit threshold; and identifying the threshold associated with the aggregate value of the items as a difference between the current total and the upper limit threshold, wherein the upper limit threshold includes a shipping value.
Veettil, on the other hand, teaches determining a current total for the items ([0018] Upon receiving the user's selection of a secondary product and a reward associated with purchase of the secondary product, the platform may display a third user interface including a basket (e.g., cart) containing the two items and a price of the basket (e.g., accounting for the selected reward); [0019] after receiving a user's selection of a secondary product (e.g., shoes) and the reward associated with the secondary product (e.g., 20% discount), the computer system can: update and/or or recalculate probabilities that the user will purchase additional, tertiary products offered by the merchant based on the selected secondary product… the computer system may repeat this process until… the basket size (e.g., price of the basket) reaches a certain threshold value);
identifying an upper limit threshold ([0019] update and/or or recalculate probabilities that the user will purchase additional, tertiary products offered by the merchant based on the selected secondary product… the computer system may repeat this process until… the basket size (e.g., price of the basket) reaches a certain threshold value); and
identifying the threshold associated with the aggregate value of the items as a difference between the current total and the upper limit threshold, wherein the upper limit threshold includes a shipping value ([0019] the computer system may repeat this process until a certain number of products are selected for purchase, a certain number of (e.g., moves) in the offer generation process are executed, or until the basket size (e.g., price of the basket) reaches a certain threshold value).
It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak, to include determining a current total for the items; identifying an upper limit threshold; and identifying the threshold associated with the aggregate value of the items as a difference between the current total and the upper limit threshold, wherein the upper limit threshold includes a shipping value, as taught by Veettil, for the same reasons discussed above with respect to claim 11.
Regarding Claim 19, Arora in view of Archak and Veettil teaches the limitations of claim 11.
Arora further discloses wherein creating the re-ranking of the ranking of the new items comprises: removing items from the ranking of the new items that are not selectable to be added to the electronic representation ([0067-0068] the complementary category filtering technique can involve applying the lift scores to the complementary items, such that truly complementary items are boosted more, while popular yet unrelated items are boosted less, such that these latter items can drop lower in the score ranking and be effectively filtered out);
removing items from the ranking of the new items that are in the electronic representation ([0093] filtering out items from the unified list… if a basket item in the basket has an L4 subcategory of “Canned Corn,” then items in the unified list that have that same L4 subcategory of “Canned Corn” can be removed from the unified list, so that the remaining items in the unified list will not be too similar to what is already in the basket); and
re-ranking, based on the ranking of the new items, remaining items of the new items ([0094] sorting each item in the unified list by the score of the item. The score of the item can be the cohesion score determined for each recommended item… adjusted based on the lift scores).
Regarding Claim 20, Arora in view of Archak and Veettil teaches the limitations of claim 11.
Arora further discloses including a first subset of the re-ranked ranking of the new items in a first portion of the GUI ([0096] the entire list of personalized item recommendations can be displayed to the user, either all at once or in portions, such as in carousels that are presented round robin to display the entire list in segments); and
including remaining ones of the re-ranked ranking of the new items in a second portion of the GUI that the user can access via interaction with the GUI ([0096] the entire list of personalized item recommendations can be displayed to the user… in portions, such as in carousels that are presented round robin to display the entire list in segments).
Regarding Claim 21, Arora discloses A non-transitory computer-readable medium storing instructions, the instructions, upon execution by a processor, cause the processor to perform operations comprising ([0021-0022]):
receiving interaction information regarding one or more interactions with a graphical user interface (GUI) of a user device ([Fig. 4]; [0046-0049] using past purchase data for users… over a one year time frame with user-item interactions; see [0096][0123] items selected for purchase via a user interface);
analyzing items in an electronic representation, for a current user session to determine a journey, using the electronic representation ([Fig. 5]; [0073-0074] receiving a basket (virtual cart) for a user including items that have been selected by the user… each item in the basket can be grouped by L3 categories);
using a machine learning model for the current user session in real-time based on the interaction information, the journey, a type corresponding to one or more types of the items, and additional information in the electronic representation to automatically determine rankings of new items to be selected, via the GUI, as additions to the electronic representation ([0042] a within-basket recommender, which can suggest grocery items that go well with the items in a shopping basket (e.g., cart) of the user, such as milk with cereals, or pasta with pasta sauce… considering both (i) item-to-item compatibility within a shopping basket and (ii) user-to-item affinity; [0051-0053] determining a top k complementary items as outputs for an anchor item j and user in order to recommend one or more items that are personalized for a user; [0109]);
creating a ranking the new items by using the machine learning model, as built ([0046] the triple embeddings model is trained to define cohesion scores; [0051-0053] the top k complementary items (i) can be determined by the triple embeddings model as outputs);
creating, based on creating the ranking of the new items using the machine learning model, a re-ranking of the ranking of the new items based on a score and a status corresponding to an item, of the new items, being selectable to be added to the electronic representation ([0067-0068] the complementary category filtering technique can involve applying the lift scores to the complementary items, such that truly complementary items are boosted more, while popular yet unrelated items are boosted less, such that these latter items can drop lower in the score ranking and be effectively filtered out; [0094] cohesion scores determined for each recommended item are adjusted based on lift scores; see [0090]); and
transmitting information based on the re-ranked ranking of the new items ([Fig. 6]; [0096] sending instructions to display at least a portion of the list of personalized recommended items to the user… the display can occur on a user interface of an electronic device).
Arora discloses analyzing items in an electronic representation, for a current user session to determine an online journey, using the electronic representation (Arora [Fig. 5][0073-0074]). However, Arora does not explicitly teach determining that an online journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected.
Archak, on the other hand, teaches determining that an online journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected ([Fig. 4]; [0032-0035]).
It would have been obvious to one of ordinary skill in the art at the time of filing to modify the machine learning-based recommendation method as taught by Arora, to include determining that an online journey is one of: routine, corresponding to first items of the items that are routinely selected; non-routine, corresponding to second items of the items that are not routinely selected; or mixed, corresponding to the first items of the items that are routinely selected and the second items of the items that are not routinely selected, as taught by Archak, for the same reasons discussed above with respect to claim 11.
Arora further discloses using a machine learning model for the current session in real-time based on the interaction information, the online journey, a type corresponding to one or more types of the items, and additional information of the items in the electronic representation to automatically determine rankings of new items to be selected, via the GUI, as additions to the electronic representation (Arora [0042][0051-0053][0109]). Arora further discloses creating a re-ranking of the ranking of the new items based on a score ([0067-0068][0090][0094]). However, Arora in view of Archak does not explicitly teach building a machine learning model based on an aggregate value of the items to determine items to be selected to satisfy the threshold associated with the aggregate value of the items; and wherein a score is a time window score.
Veettil, on the other hand, teaches building a machine learning model based on an aggregate value of the items to determine items to be selected to satisfy the threshold associated with the aggregate value of the items ([0052] generating a user-specific intent model to intake a set of basket characteristics (e.g., a price associated with the primary product)… as a user selects rewards offered to her over time, the computer system can record product characteristics (types, prices, promotional discounts, product groupings); see [0019] displaying purchase options to a user in response to a user’s selection of a product and the reward associated with the product (e.g., 20% discount)… the offer generation process is executed repeatedly until the basket size (e.g., the price of the basket) reaches a certain threshold value); and
wherein a score is a time window score ([0052] the intent model intakes a set of basket characteristics (e.g., a time of day) and outputs a predicted intent score for purchase of a particular secondary product).
It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak, to include building a machine learning model based on an aggregate value of the items to determine items to be selected to satisfy the threshold associated with the aggregate value of the items; and wherein a score is a time window score, as taught by Veettil, for the same reasons discussed above with respect to claim 11.
Regarding Claim 23, Arora in view of Archak and Veettil teaches the limitations of claim 21.
Arora further discloses wherein creating the ranking of the new items comprises: removing items, from the new items, to obtain a reduced set of the new items ([0067-0068] the complementary category filtering technique can involve applying the lift scores to the complementary items, such that truly complementary items are boosted more, while popular yet unrelated items are boosted less, such that these latter items can drop lower in the score ranking and be effectively filtered out); and
ranking the reduced set of the new items by using the machine learning model after removing the items from the new items ([0094] sorting each item in the unified list by the score of the item. The score of the item can be the cohesion score determined for each recommended item… adjusted based on the lift scores).
Regarding Claim 24, Arora in view of Archak and Veettil teaches the limitations of claim 21.
Arora further discloses wherein using the machine learning model comprises: using, to improve operations of a serving layer, a backend server to train the machine learning model ([Fig. 7]; [0107-0109] the system performs offline training periodically to handle new past-purchase transaction data to update the model; see [Fig. 1][0028] computer system 100 may comprise a single server). However, Arora does not explicitly teach building the machine learning model.
Veettil, on the other hand, teaches building the machine learning model ([0052] generating a user-specific intent model to intake a set of basket characteristics (e.g., a price associated with the primary product)).
It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak, to include building the machine learning model, as taught by Veettil, for the same reasons discussed above with respect to claim 21.
Regarding Claim 26, Arora in view of Archak and Veettil teaches the limitations of claim 21.
Arora further discloses using a backend server to train the machine learning model ([Fig. 7]; [0107-0109] the system performs offline training periodically to handle new past-purchase transaction data to update the model; see [Fig. 1][0028] computer system 100 may comprise a single server).
Claims 15 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Arora in view of Archak and Veettil, and further in view of previously cited Trepca (US 2016/0140519 A1).
Regarding Claim 15, Arora in view of Archak and Veettil teaches the limitations of claim 11.
Arora further discloses wherein using the machine learning model for the current user session in real-time comprises: building a matrix based on the interaction information ([0045] The triple embeddings model thus uses triplets of (user, first item, second item), indicating that the first and second items were bought by the user in the same basket; [0048] the triple embeddings model can be trained for the two sets of item embeddings (p, q) and the user embeddings (h)… the triple embeddings model can be trained… using 500 million triplets, using a past purchase data set over a one year time frame with 800 million user-item interactions; [0049] matrix P and matrix Q can store the two sets of trained item embeddings for the catalog of items… Matrix H can store the trained user embeddings for the users); and
using a model to determine one or more of user embeddings or item embeddings ([0046] the triple embeddings model can be trained using past purchase data for users to derive embeddings that represent the users and the items from the triplets).
Veettil, on the other hand, teaches building the model ([0019][0052]).
It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak, to include building the model, as taught by Veettil, for the same reasons discussed above with respect to claim 11.
While Arora discloses building a matrix based on the interaction information ([0045][0048-0049]), and using a model to determine one or more of user embeddings or item embeddings ([0046]), and Veettil teaches building the model ([0019][0052]), Arora in view of Archak and Veettil does not explicitly teach wherein a matrix is a user-interaction matrix; and building a sensitive Weighted Alternate Least Squares (WALS) model.
However, in the field of item recommendations (see at least Trepca [0069-0077]), Trepca, on the other hand, teaches wherein a matrix is a user-interaction matrix ([0428] the input data of the algorithm consists of a user-product matrix: a binary (0/1) matrix in which rows represented users, and column represented products; positive entries in the matrix represented interactions between a given user and product, while zero entries denoted the lack of any interaction); and
building a sensitive Weighted Alternate Least Squares (WALS) model ([0427] User and product models are computed using a modified version of the Weighted Alternating Least Squares (WALS) algorithm; [0434] To surmount these difficulties, a modified version of the WALS algorithm is used, which performs joint estimation of user latent factors and content coefficients, i.e. user preferences for price).
The steps of Trepca are applicable to the method of Arora in view of Archak and Veettil, as they share characteristics and capabilities, namely, they are directed to providing item recommendations based on user selections. It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak and Veettil, to include wherein a matrix is a user-interaction matrix; and building a sensitive Weighted Alternate Least Squares (WALS) model, as taught by Trepca. One of ordinary skill in the art at the time of filing would have been motivated to expand the recommendation method of Arora in view of Archak and Veettil in order to combine the advantages of content-based and collaborative filtering recommender systems, and allow prediction of user preference for new items (Trepca, [0445]).
Regarding Claim 17, Arora in view of Archak and Veettil teaches the limitations of claim 11.
Arora further discloses identifying an output from a model, the output including a ranked list of the new items ([0046] the triple embeddings model can be trained using past purchase data for users to derive embeddings that represent the users and the items from the triplets… These embeddings can be modeled… such that a cohesion score… can be defined; [0052] for the anchor item j and user u, given as inputs, the top k complementary items (i) can be determined as outputs by iterating through the items (i) in matrix P and computing the cohesion score, and selecting the top k items (i));
determining a filtered list of new items by removing a number of the new items ([0053] the item-to-item model additionally can include a complementary category filtering technique, which can filter out items that are recommended due to being popular overall items… These items would often be included as recommendations from the item-to-item model, merely due to their popularity in most carts, despite not being particularly complementary to a given anchor item j; [0067] the complementary category filtering technique can involve applying the lift scores to the complementary items, such that truly complementary items are boosted more, while popular yet unrelated items are boosted less, such that these latter items can drop lower in the score ranking and be effectively filtered out); and
identifying a threshold number of items from each type in the filtered list of the new items ([0078] As an example, there can be 8 items in the basket that was selected by a user… If the total number of item recommendations that will be generated for the basket is 40, then the number of item recommendations sampled for each of the first two L3 categories can be 15, which is ⅜ of 40, and the number of item recommendations sampled for each of the last two L3 categories can be 5, which is ⅛ of 40. In another embodiment, the number k in the request for the top k items requested in each of blocks 531-534 can be varied based on the proportion of items in each L3 category).
However, Arora does not explicitly teach wherein a model is a sensitive WALS model.
Trepca, on the other hand, teaches wherein a model is a sensitive WALS model ([0427] User and product models are computed using a modified version of the Weighted Alternating Least Squares (WALS) algorithm; [0434] To surmount these difficulties, a modified version of the WALS algorithm is used, which performs joint estimation of user latent factors and content coefficients, i.e. user preferences for price).
It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak and Veettil, to include wherein a model is a sensitive WALS model, as taught by Trepca, for the same reasons discussed above with respect to claim 15.
Claim 25 is rejected under 35 U.S.C. 103 as being unpatentable over Arora in view of Archak and Veettil, and further in view of newly cited U.S Patent Application No. 2022/0277375 A1 to Su et al, hereinafter Su.
Regarding Claim 25, Arora in view of Archak and Veettil teaches the limitations of claim 21.
Arora further discloses wherein creating the ranking of the new items comprises: reducing a size of a set of the new items ([0067-0068] the complementary category filtering technique can involve applying the lift scores to the complementary items, such that truly complementary items are boosted more, while popular yet unrelated items are boosted less, such that these latter items can drop lower in the score ranking and be effectively filtered out); and
using the machine learning model to create the ranking of the new items after reducing the size of the set of the new items ([0094] sorting each item in the unified list by the score of the item. The score of the item can be the cohesion score determined for each recommended item in block 625, which in some embodiments, was adjusted based on the lift scores). However, Arora in view of Archak and Veettil does not explicitly teach reducing a size of a set of the new items before using the machine learning model.
However, in the field of item recommendations during a web session (see at least Se [abstract][0028-0030]), Su, on the other hand, teaches reducing a size of a set of the new items before using the machine learning model ([0049] search engine 118 uses the data specified in the search query to identify the item listings stored in the item data storage device 124 in response to the search query; see [0029-0030] retrieving listing items matching a search query; [0050-0051])
The steps of Su are applicable to the method of Arora in view of Archak and Veettil, as they share characteristics and capabilities, namely, they are directed to providing item recommendations during a web session. It would have been obvious to one of ordinary skill in the art at the time of filing to modify the recommendation method as taught by Arora in view of Archak and Veettil, to include reducing a size of a set of the new items before using the machine learning model, as taught by Su. One of ordinary skill in the art at the time of filing would have been motivated to expand the recommendation method of Arora in view of Archak and Veettil in order to provide a recommendation system capable of reacting to a user’s changing preferences within a web session (Su, [0445]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ZACHARY R DONAHUE whose telephone number is (571)272-5850. The examiner can normally be reached M-F 8a-5p.
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, Marissa Thein can be reached at (571) 272-6764. 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.
/ZACHARY RYAN DONAHUE/ Examiner, Art Unit 3689
/MARISSA THEIN/ Supervisory Patent Examiner, Art Unit 3689
Prosecution Insights