DETAILED ACTION
Continued Examination under 37 CFR 1.114
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 on 03/27/2026 has been entered.
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 .
Examiner’s Comment
This Action is in response to the Request for Continued Examination filed on 03/27/2026 with Amended Claims and Applicant's Remarks filed on 03/27/2026.
Applicant has amended claims 2, 5, 9, 11, 12, 14, 16, 17, and 19 according to Amendments filed on 03/27/2026. Claims 2-20 are pending and currently under consideration for patentability.
Information Disclosure Statement
The information disclosure statement(s) (IDS) submitted on 03/27/2026 has/have been considered by the examiner.
Double Patenting
The nonstatutory double patenting rejection is based on a judicially created doctrine grounded in public policy (a policy reflected in the statute) so as to prevent the unjustified or improper timewise extension of the “right to exclude” granted by a patent and to prevent possible harassment by multiple assignees. A nonstatutory double patenting rejection is appropriate where the conflicting claims are not identical, but at least one examined application claim is not patentably distinct from the reference claim(s) because the examined application claim is either anticipated by, or would have been obvious over, the reference claim(s). See, e.g., In re Berg, 140 F.3d 1428, 46 USPQ2d 1226 (Fed. Cir. 1998); In re Goodman, 11 F.3d 1046, 29 USPQ2d 2010 (Fed. Cir. 1993); In re Longi, 759 F.2d 887, 225 USPQ 645 (Fed. Cir. 1985); In re Van Ornum, 686 F.2d 937, 214 USPQ 761 (CCPA 1982); In re Vogel, 422 F.2d 438, 164 USPQ 619 (CCPA 1970); In re Thorington, 418 F.2d 528, 163 USPQ 644 (CCPA 1969).
A timely filed terminal disclaimer in compliance with 37 CFR 1.321(c) or 1.321(d) may be used to overcome an actual or provisional rejection based on nonstatutory double patenting provided the reference application or patent either is shown to be commonly owned with the examined application, or claims an invention made as a result of activities undertaken within the scope of a joint research agreement. See MPEP § 717.02 for applications subject to examination under the first inventor to file provisions of the AIA as explained in MPEP § 2159. See MPEP §§ 706.02(l)(1) - 706.02(l)(3) for applications not subject to examination under the first inventor to file provisions of the AIA . A terminal disclaimer must be signed in compliance with 37 CFR 1.321(b).
The USPTO Internet website contains terminal disclaimer forms which may be used. Please visit www.uspto.gov/patent/patents-forms. The filing date of the application in which the form is filed determines what form (e.g., PTO/SB/25, PTO/SB/26, PTO/AIA /25, or PTO/AIA /26) should be used. A web-based eTerminal Disclaimer may be filled out completely online using web-screens. An eTerminal Disclaimer that meets all requirements is auto-processed and approved immediately upon submission. For more information about eTerminal Disclaimers, refer to www.uspto.gov/patents/process/file/efs/guidance/eTD-info-I.jsp.
Claims 2-20 are rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-25 of U.S. Patent No. 11,068,935 because the patent and the application under examination name the same inventive entity. Although the claims at issue are not identical, they are not patentably distinct from each other because the claims in Patent No. 11,068,935 recite the entirety of limitations of claims 2 and 11 of the instant application. For example, application claims 1 and 11 are anticipated by patent claims 1, 7, and 19 because patent claims 1, 7, and 19 recite additional features such as “generating a p-dimensional embedding of a plurality of websites based on the first plurality website visitation records; accessing a plurality of conversion event data, the plurality of conversion event data associated with a second plurality website visitation records, the plurality of conversion event data indicating that a subset of the second plurality of user devices that has not opted out of having conversion event data tracked performed a conversion action, each website visitation record from the second plurality of website visitation records associated with a user device from a second plurality of user devices and indicating a plurality of websites visited by that user device; determining, for each website visitation record from the second plurality of website visitation records, a position of each website from the plurality of websites indicated in that website visitation record in the p-dimensional embedding; receiving an indication that a user device is accessing a website, the user device that accessed the website not being from the first plurality of user devices or the second plurality of user devices, cookie-based tracking information for the user device being disabled; determining the position of the website in the p-dimensional embedding; and facilitating delivery of targeted content to the user device that accessed the website based on predicting, using the machine learning model, a likelihood of whether the user device that accessed the website will perform a conversion action based on the position of the website in the p-dimensional embedding and without accessing cookie-based tracking information for the user device” wherein application claims 2 and 11 do not recite these features and are essentially broader than patent claims 1, 7, and 19. Therefore patent claims 1, 7, and 19 of Patent No. 11,068,935 is in essence a “species” of the generic invention of application claims 2 and 11. It has been held that a generic invention is “anticipated” by a “species” within the scope of the generic invention. See In re Goodman, 29 USPQ2d 2010 (Fed. Cir. 1993). Claims 3-10 (Dependent on Claim 2) and claims 12-20 (Dependent on Claim 11) do not cure the deficiencies of the independent claims. Appropriate correction is required.
Claims 2-20 are rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-21 of U.S. Patent No. 11,699,109 because the patent and the application under examination name the same inventive entity. Although the claims at issue are not identical, they are not patentably distinct from each other because the claims in Patent No. 11,699,109 recite the entirety of limitations of claims 2 and 11 of the instant application. For example, application claims 2 and 11 are anticipated by patent claims 1 and 20 because patent claims 1 and 20 recite additional features such as “receiving an indication that a user device that does not report cookie information is accessing a website; defining an encoding layer that represents a plurality of websites including the website based on web browsing history data from a plurality of user devices that has not opted out of having website browsing history tracked, the user device not being included in the plurality of user devices, the encoding layer being a reduced dimensional space capturing similarities between websites from the plurality of websites” wherein application claims 2 and 11 do not recite these features and are essentially broader than patent claims 1 and 20. Therefore patent claims 1 and 20 of Patent No. 11,699,109 is in essence a “species” of the generic invention of application claims 2 and 11. It has been held that a generic invention is “anticipated” by a “species” within the scope of the generic invention. See In re Goodman, 29 USPQ2d 2010 (Fed. Cir. 1993). Claims 3-10 (Dependent on Claim 2) and claims 12-20 (Dependent on Claim 11) do not cure the deficiencies of the independent claims. Appropriate correction is required.
Claims 2-20 are rejected on the ground of nonstatutory double patenting as being unpatentable over claims 1-29 of U.S. Patent No. 12,045,703 because the patent and the application under examination name the same inventive entity. Although the claims at issue are not identical, they are not patentably distinct from each other because the claims in Patent No. 12,045,703 recite the entirety of limitations of claims 2 and 11 of the instant application. For example, application claims 2 and 11 are anticipated by patent claims 1 and 24 because patent claims 1 and 24 recite additional features such as “accessing a history of uniform resource locators (URLs) visited by each user device from a first group of user devices; generating an embedding of a first plurality of websites based on the history of URLs visited by the first group of user devices; the conversion event data including no data tied to user identifiers associated with the user devices of the second group of user devices, the second plurality of websites at least partially overlapping with the first plurality of websites; and training a machine learning model using the embedding and the conversion event data such that, given a position in the embedding, including positions in the embedding that are not associated with the second plurality of websites and for which conversion event data is not available, the machine learning model is configured to predict a likelihood that an item of targeted content will produce a conversion event, the position in the embedding associated with a website accessed by a user device to which the item of targeted content is presented and having a privacy setting configured such that the user device does not provide a tracking identifier operable to identify a history of URLs visited by the user device” wherein application claims 2 and 11 do not recite these features and are essentially broader than patent claims 1 and 24. Therefore patent claims 1 and 24 of Patent No. 12,045,703 is in essence a “species” of the generic invention of application claims 2 and 11. It has been held that a generic invention is “anticipated” by a “species” within the scope of the generic invention. See In re Goodman, 29 USPQ2d 2010 (Fed. Cir. 1993). Claims 3-10 (Dependent on Claim 2) and claims 12-20 (Dependent on Claim 11) do not cure the deficiencies of the independent claims. Appropriate correction is required.
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 2-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. The claims are directed to a judicial exception (i.e., a law of nature, natural phenomenon, or abstract idea) without significantly more.
Step 1: In a test for patent subject matter eligibility, claims 2-20 are found to be in accordance with Step 1 (see 2019 Revised Patent Subject Matter Eligibility), as they are related to a process, machine, manufacture, or composition of matter. Claims 2-20 recite a computer-readable medium. When assessed under Step 2A, Prong I, they are found to be directed towards an abstract idea. The rationale for this finding is explained below:
Step 2A, Prong I: Under Step 2A, Prong I, claims 2 and 11 are directed to an abstract idea without significantly more, as they all recite a judicial exception. Claims 2 and 11 recite limitations directed to the abstract idea including access a word embedding; determine a location in the word embedding associated with at least one of each website from a plurality of websites or each audience from a plurality of audiences, a location of each website from the plurality of websites determined in the word embedding based on visitation data associated with that website and/or a location of each audience from the plurality of audiences determined in the word embedding based on behavioral data for a user associated with that audience; identify, for a subset of conversion actions taken, a corresponding website visited; receive a search query associated with an item of targeted content; determine a search query location in the word embedding based on the search query; identify at least one of a target audience from the plurality of audiences or a target website from the plurality of websites based on a distance, in the word embedding, between the search query location and a location of the at least one of the target audience or the target website; and send a signal to facilitate delivery of the item of targeted content to at least one of the target website or an identified user associated with the target audience based on the predicted likelihood of conversion for the at least one of the target audience or the target website. These further limitations are not seen as any more than the judicial exception. Claims 2 and 11 recite additional limitation “a non-transitory, processor-readable medium storing code representing instructions to be executed by a processor; associated with devices that have behavioral-based tracking enabled; and from a device that has cookie-based tracking disabled”. Claim 2 recites additional limitations including “define training data based on determining a location in the word embedding for each website visited that is associated with a conversion action taken; train a machine learning model using the training data to predict a likelihood of conversion based on a location in the word embedding; and apply the machine learning model to the location in the word embedding of the at least one of the target audience or the target website to predict a likelihood of conversion for the at least one of the target audience or the target website”. Facilitating delivery of targeted content based on a word embedding or representation of website visitation data and behavioral data is considered to be an abstract idea, specifically, certain methods of organizing human activity; such as commercial interactions, advertising, marketing, and sales because the claims are directed to facilitating delivery of targeted content (i.e. advertising) based on website visitation data and behavioral data associated with conversion event (i.e. commercial interaction) which is a basic commerce problems merchants’ try to solve. Furthermore, the analysis in which identifying the target audience takes place falls under another abstract idea, specifically mental processes; such as concepts performed in the human mind (including an observation, evaluation, judgement, opinion) because the claims are directed to such limitations as accessing data (i.e. word embedding of website visitation data and behavioral data associated with conversion event) and receiving data (i.e. search query) in order to train a machine learning model. Therefore, under Step 2A, Prong I, claims 2 and 11 are directed towards an abstract idea.
Step 2A, Prong II: Step 2A, Prong II is to determine whether any claim recites any additional element that integrate the judicial exception (abstract idea) into a practical application. Claims 2 and 11 recite additional limitation “a non-transitory, processor-readable medium storing code representing instructions to be executed by a processor; associated with devices that have behavioral-based tracking enabled; and from a device that has cookie-based tracking disabled”. Claim 2 recites additional limitations including “define training data based on determining a location in the word embedding for each website visited that is associated with a conversion action taken; train a machine learning model using the training data to predict a likelihood of conversion based on a location in the word embedding; and apply the machine learning model to the location in the word embedding of the at least one of the target audience or the target website to predict a likelihood of conversion for the at least one of the target audience or the target website”. Claims 2 and 11 reciting additional limitations including “a non-transitory, processor-readable medium storing code representing instructions to be executed by a processor; associated with devices that have behavioral-based tracking enabled; and from a device that has cookie-based tracking disabled” do not integrate the claims into practical application. Merely reciting the processor-readable medium and devices that have behavioral-based tracking enabled/disabled in this manner is seen as adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f). Accordingly, alone, and in combination, these additional elements are seen as using a computer or tool to perform an abstract idea, adding insignificant-extra-solution activity to the judicial exception. They do no more than link the judicial exception to a particular technological environment or field of use, i.e. processor-readable medium and devices, and therefore do not integrate the abstract idea into a practical application. The courts decided that although the additional elements did limit the use of the abstract idea, the court explained that this type of limitation merely confines the use of the abstract idea to a particular technological environment and this fails to add an inventive concept to the claims (See Affinity Labs of Texas v. DirecTV, LLC,). Under Step 2A, Prong II, these claims remain directed towards an abstract idea.
Step 2B: Claims 2 and 11 recite additional limitation “a non-transitory, processor-readable medium storing code representing instructions to be executed by a processor; associated with devices that have behavioral-based tracking enabled; and from a device that has cookie-based tracking disabled”. Claim 2 recites additional limitations “define training data based on determining a location in the word embedding for each website visited that is associated with a conversion action taken; train a machine learning model using the training data to predict a likelihood of conversion based on a location in the word embedding; and apply the machine learning model to the location in the word embedding of the at least one of the target audience or the target website to predict a likelihood of conversion for the at least one of the target audience or the target website”. Claims 2 and 11 reciting the following additional elements: “a non-transitory, processor-readable medium storing code representing instructions to be executed by a processor; associated with devices that have behavioral-based tracking enabled; and from a device that has cookie-based tracking disabled” do not integrate the judicial exception (abstract idea) into a practical application because of the analysis provided in Step 2A, Prong II. Claim 2 also recites the limitation - “define training data based on determining a location in the word embedding for each website visited that is associated with a conversion action taken; train a machine learning model using the training data to predict a likelihood of conversion based on a location in the word embedding; and apply the machine learning model to the location in the word embedding of the at least one of the target audience or the target website to predict a likelihood of conversion for the at least one of the target audience or the target website”. However, these additional elements or a combination of elements do not result in the claims amounting to significantly more than the judicial exception because they are not indicative of integration into a practical application. Merely training a machine learning model with known inputs in a specific format (e.g. word embedding representing received conversion event data), applying the inputs to the machine learning model to output data in a specific format (i.e. vector representation of websites) in order to predict a likelihood of an event (e.g. conversion) is seen as adding insignificant extra-solution activity to the judicial exception - see MPEP 2106.05(g). It has been well-known since at least 1996 that the “Machine learning algorithms build a model based on sample data, known as "training data", in order to make predictions or decisions without being explicitly programmed to do so.” (See Wikipedia: Machine learning: The definition "without being explicitly programmed" is often attributed to Arthur Samuel, who coined the term "machine learning" in 1959, but the phrase is not found verbatim in this publication, and may be a paraphrase that appeared later. Confer "Paraphrasing Arthur Samuel (1959), the question is: How can computers learn to solve problems without being explicitly programmed?" in Koza, John R.; Bennett, Forrest H.; Andre, David; Keane, Martin A. (1996). Automated Design of Both the Topology and Sizing of Analog Electrical Circuits Using Genetic Programming. Artificial Intelligence in Design '96. Springer, Dordrecht. pp. 151–170. doi:10.1007/978-94-009-0279-4_9.”). Claims 2 and 11 do not include additional elements or a combination of elements that result in the claims amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional elements listed amount to no more than mere instructions to apply an exception using a generic computer component. In addition, the applicant’s specifications describe “general purpose processor” elements, ¶ [0031], for implementing the user device, which do not amount to significantly more than the abstract idea of itself, which is not enough to transform an abstract idea into eligible subject matter. Furthermore, there is no improvement in the functioning of the computer or technological field, and there is no transformation of subject matter into a different state. Under Step 2B in a test for patent subject matter eligibility, these claims are not patent eligible.
Dependent claims 3-10; and 12-20 further recite the computer-readable medium of claims 2 and 11, respectively. Dependent claims 3-10 and 12-20 when analyzed as a whole are held to be patent ineligible under 35 U.S.C. 101 because the additional recited limitation fail to establish that the claims are not directed to an abstract idea:
Under Step 2A, Prong I, these additional claims only further narrow the abstract idea set forth in claims 2 and 11. For example, claims 3-10 and 12-20 describe the limitations for facilitating delivery of targeted content based on a word embedding or representation of website visitation data and behavioral data – which is only further narrowing the scope of the abstract idea recited in the independent claims.
Under Step 2A, Prong II, for dependent claims 3-10 and 12-20, there are no additional elements introduced. Thus, they do not present integration into a practical application, or amount to significantly more.
Under Step 2B, the dependent claims do not include any additional elements that are sufficient to amount to significantly more than the judicial exception. Additionally, there is no improvement in the functioning of the computer or technological field, and there is no transformation of subject matter into a different state. As discussed above with respect to integration of the abstract idea into a practical application, the additional claims do not provide any additional elements that would amount to significantly more than the judicial exception. Under Step 2B, these claims are not patent eligible.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 2-20 are rejected under 35 U.S.C. 103 as being unpatentable over U.S. Publication 2015/0379571 to Grbovic in view of U.S. Publication 2019/0197398 to Jamali and in further view of U.S. Publication 2014/0297394 to Li.
With respect to Claim 1:
Grbovic teaches:
A non-transitory, processor-readable medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to (Grbovic: ¶ [0045]):
access a word embedding (i.e. access word representation sessionalized for user segment or user profile type including plurality of audiences or user segments) (Grbovic: ¶ [0079] “At block 414, the system circuitry applies one or more modified linguistic modeling or statistical natural language processing techniques, such as a modified skip-gram model in some embodiments, to the results of the pre-processing in order to identify distributed query word representations in the historical web search data. In some embodiments, the distributed query word representations consist of associations between search query terms and ad clicks to the actions of a user. For example, the distributed query word representations may represent a likelihood that a user will perform a given action ( e.g., click on a displayed ad related to a particular category) after the user enters a search query containing a particular keyword. Traditional natural language processing techniques may typically involve one or more algorithms performed on an article, set of articles, or similar body of text that are input into to the algorithm and treated as "documents." Each "word" in the document is then analyzed to determine the statistical relevance.”);
determine a location in the word embedding associated with at least one of each website from a plurality of websites or each audience from a plurality of audiences associated with devices that have behavioral-based tracking enabled, a location of each website from the plurality of websites determined in the word embedding based on visitation data associated with that website and/or a location of each audience from the plurality of audiences determined in the word embedding based on behavioral data for a user associated with that audience (i.e. determine location in word representation sessionalized for user segment or user profile type including plurality of audiences or user segments, and the word representation includes keywords/query terms that lead user throughout a path within a webpage that may lead to a conversion, wherein the user devices allow for user behavior such as actions by the user to be tracked) (Grbovic: ¶ [0079] “At block 414, the system circuitry applies one or more modified linguistic modeling or statistical natural language processing techniques, such as a modified skip-gram model in some embodiments, to the results of the pre-processing in order to identify distributed query word representations in the historical web search data. In some embodiments, the distributed query word representations consist of associations between search query terms and ad clicks to the actions of a user. For example, the distributed query word representations may represent a likelihood that a user will perform a given action ( e.g., click on a displayed ad related to a particular category) after the user enters a search query containing a particular keyword. Traditional natural language processing techniques may typically involve one or more algorithms performed on an article, set of articles, or similar body of text that are input into to the algorithm and treated as "documents." Each "word" in the document is then analyzed to determine the statistical relevance.” Furthermore, as cited in ¶ [0040] “Another approach includes profile-type ad targeting. In this approach, user or group profiles specific to a respective user or group may be generated to model user behavior, for example, by tracking a user's path through a website or network of sites, and compiling a profile based, at least in part, on ad GUIs, webpages, and advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users. The aforementioned profile-type targeting data may be logged in data logs and such logs may be communicated to the analytics server 118 for processing. Once processed into corresponding analytics data, the analytics server 118 can provide analyzed feedback for affecting future serving of content.” Furthermore, as cited in ¶¶ [0093] [0094] “Traditional computational linguistic techniques will typically consider words associations within a text-based document without consideration of whether the term comes before or after the word being examined in order to determine the relevance of the words to each other. In other words, the elements of a document are not treated differently for analytical approach based on their location within the document. However, in web search activity, the primary focus is on the data immediately preceding an ad click as, conceptually, this is most likely to be representative of why the user clicked on the ad. Thus, some embodiments further adapt the skip-gram modeling techniques to make the process directed such that it considers only the preceding actions within a certain distance the ad click. While this approach would not make sense in a traditional skip-gram modeling, the modification results in improved distributed representations for web search data due in part to the unique nature of web search activity…Additionally, in some embodiments, the web search activity may be weighted based on recency or distance in time from a particular ad click.”);
identify, for a subset of conversion actions taken, a corresponding website visited (i.e. identify conversion event data includes click data occurring at website) (Grbovic: ¶ [0083] “At block 518, the system circuitry retrieves historical web search data related the identified ad categories from the system databases. By way of illustration, the historical web search data may include historical search queries entered by users, historical advertising campaigns, recorded ad clicks or interactions with ad content, ad impressions, and resulting ad conversions, for example…The activity for each user is recorded as one record in the activity logs. The system may retrieve all web search data for a recent period of time, such as for the past six months, and examine the data on a per-user basis to determine keyword relevancy to the particular user. In this way, the system can ultimately generated targeted keyword lists for a particular user, or set of users determined to be similar based on known profile traits, in order to provide search retargeting rules that target the particular user or set of users having similar traits.”);
define training data based on determining a location in the word embedding for each website visited that is associated with a conversion action taken (i.e. define training data to be query words in vector space that lead user through navigation to click) (Grbovic: ¶ [0023] “In particular, certain embodiments are directed systems and methods for generating data-driven keyword clusters using distributed query word representations formed from novel techniques for analyzing and processing historical web search activity, including, by way of illustration, historical search queries entered by users, historical advertising campaigns, recorded ad clicks or interactions, ad impressions, and resulting ad conversions, for example. Keyword cluster sets or keyword lists for a specific advertiser or campaign type may be generated by learning distributed representations of user queries that are most likely to lead to ad clicks and conversions. In some embodiments, the distributed representation may be generated by applying a directed approach to learning distributed representations that focuses on or weights the data to emphasize actions immediately preceding an ad click. For example, by using deep learning technologies, the circuitry components of the present system generate distributed representations of query words in vector space using the search engine data, such that similar words in context of web search (i.e., those that are most likely to lead to ad clicks) can be found in a cluster K of the nearest neighbors of an adword or keyword category.”);
train a machine learning model using the training data to predict a likelihood of conversion based on a location in the word embedding (i.e. training deep learning model using word representation and conversion event data in order to predict a likelihood that a content will receive a click or conversion) (Grbovic: ¶ [0023] “In particular, certain embodiments are directed systems and methods for generating data-driven keyword clusters using distributed query word representations formed from novel techniques for analyzing and processing historical web search activity, including, by way of illustration, historical search queries entered by users, historical advertising campaigns, recorded ad clicks or interactions, ad impressions, and resulting ad conversions, for example. Keyword cluster sets or keyword lists for a specific advertiser or campaign type may be generated by learning distributed representations of user queries that are most likely to lead to ad clicks and conversions. In some embodiments, the distributed representation may be generated by applying a directed approach to learning distributed representations that focuses on or weights the data to emphasize actions immediately preceding an ad click. For example, by using deep learning technologies, the circuitry components of the present system generate distributed representations of query words in vector space using the search engine data, such that similar words in context of web search (i.e., those that are most likely to lead to ad clicks) can be found in a cluster K of the nearest neighbors of an adword or keyword category.” Furthermore, as cited in ¶ [0066] “In some embodiments, modeling circuitry 204 uses computational linguistic analysis techniques that utilize aspects of skip-gram modeling to process web search activity. Instead of processing word and documents, the modified modeling program processes historical web search activity, treating ad clicks and search queries in a manner akin to how one may treat words of a document in linguistic analysis. The modeling techniques are further adapted to consider time-related data associated with the web search activity, such that the algorithm is time-sensitive. In this way, the system circuitry, including modeling circuitry 204, can generate vector representations of keywords that are statistically indicative of the correlation between ad clicks, search query terms, and targeting keywords. In other words, modeling circuitry 204 generates vector representations of the likelihood that a keyword is related to a category of an advertisement that the keyword is likely to lead to an ad click.”);
receive, [[from a device that has cookie-based tracking disabled]], a search query associated with an item of targeted content (i.e. receive search query) (Grbovic: ¶ [0064] “FIG. 2 illustrates a block diagram of circuitry components of a sponsored verb generator according to some embodiments. Keyword vector generator 200 may be communicatively coupled search retargeting framework server 116 and may include retargeting circuitry 202, modeling circuitry 204, training circuitry 206, and/or display logic circuitry 208 components. Search retargeting framework server 116 may receive a search query to from a user device and determine one or more search suggestions, sponsored or non-sponsored search results, advertisements, or other related ad content to display to the user.”);
determine a search query location in the word embedding based on the search query (i.e. determine search query location in vector embedding based on search query) (Grbovic: ¶ [0064] “For each search query entered by the user, search retargeting framework server 116 may seek to identify opportunities for monetization, including by using search retargeting rules for keyword lists that have been generated by keyword vector generator 200 using directed distributed query word representations. Search retargeting framework server 116 will communicate the requests containing search query words to keyword vector generator 200. The request will be received by keyword vector generator 200 and retargeting circuitry 202 will determine one or retargeting rules to be used in selecting an advertisement or sponsored content for display to the user request.” Furthermore, as cited in ¶¶ [0079] [0080] “For each search query entered by the user, search retargeting framework server 116 may seek to identify opportunities for monetization, including by using search retargeting rules for keyword lists that have been generated by keyword vector generator 200 using directed distributed query word representations. Search retargeting framework server 116 will communicate the requests containing search query words to keyword vector generator 200. The request will be received by keyword vector generator 200 and retargeting circuitry 202 will determine one or retargeting rules to be used in selecting an advertisement or sponsored content for display to the user request…In some embodiments, the result of steps 414 and 416 is in the form of a vector representing the keyword distributions as related to the input category or advertiser. In these embodiments, as well as others, the keywords that are most closely related to the input advertiser name or ad category are represented in the vector as being nearest to the advertiser name or ad category. In this way, the set of the most closely related keywords in the vector representations can be selected as having the highest likelihood that they are indicative or predictive of an ad click. In addition, in some embodiments, the system circuitry may generate a set of retargeting rules using the keyword list and the closest K neighbors in the list to be used in conjunction with search retargeting techniques.”);
identify at least one of […] a target website from the plurality of websites based on a distance, in the word embedding, between the search query location and a location of the at least one of the target audience or the target website (i.e. presenting search results based on distance in the embedding between search query and web search activity, wherein the search results present target websites) (Grbovic: ¶ [0064] “For each search query entered by the user, search retargeting framework server 116 may seek to identify opportunities for monetization, including by using search retargeting rules for keyword lists that have been generated by keyword vector generator 200 using directed distributed query word representations. Search retargeting framework server 116 will communicate the requests containing search query words to keyword vector generator 200. The request will be received by keyword vector generator 200 and retargeting circuitry 202 will determine one or retargeting rules to be used in selecting an advertisement or sponsored content for display to the user request. The advertisement or sponsored content components may include one or more subGUIs that are generated by or associated with the search result generated by the search result circuitry, such as various circuitry components of the search result circuitry framework 610 and described in connection with FIG. 6. The search result circuitry framework (e.g., search suggestion circuitry, webpage search result circuitry, configuration circuitry, analytics circuitry, monetization circuitry, maps circuitry, social media circuitry, and retargeting campaign generator) will generate search result content to display to the user. User interactions with the search result content, including ad impressions and ad clicks, are stored by the search retargeting server 116 and communicated to the keyword vector generator 200.” Furthermore, as cited in ¶ [0101] “The components of search result circuitry 610 may generate search results related to the search query term. As part of this process, the search suggestion circuitry 622 may generate search suggestions related to the search query to display interleaved with the search results generated by webpage search result circuitry 624. The ordering and layout of the search results and suggestions, as well as other elements on the page, may be generated by configuration circuitry 626 and may consider user profile attributes and preferences retrieved from a user profile related to the user that submitted the search query using device 601. As part of the search results, one or more map features may be generated by maps circuitry 630.”);
apply the machine learning model to the location in the word embedding of the at least one of the target audience or the target website to predict a likelihood of conversion for the at least one of the target audience or the target website (i.e. applying machine learning or deep learning to keyword locations in the word embedding of the target website or user in order to determine a predicted likelihood that a conversion will occur) (Grbovic: ¶ [0023] “Keyword cluster sets or keyword lists for a specific advertiser or campaign type may be generated by learning distributed representations of user queries that are most likely to lead to ad clicks and conversions. In some embodiments, the distributed representation may be generated by applying a directed approach to learning distributed representations that focuses on or weights the data to emphasize actions immediately preceding an ad click. For example, by using deep learning technologies, the circuitry components of the present system generate distributed representations of query words in vector space using the search engine data, such that similar words in context of web search (i.e., those that are most likely to lead to ad clicks) can be found in a cluster K of the nearest neighbors of an adword or keyword category.” Furthermore, as cited in ¶ [0114] “Other aspects may also affect ranking, such as, for example, proximity of query terms within a located record or file, or semantic usage, for example. A score and an identifier for a located record or file, for example, may be stored in a respective entry of a ranking list. A list of search results may be ranked in accordance with scores, which may, for example, be provided in response to a search query. In some embodiments, machine-learned ranking (MLR) models are used to rank search results. MLR is a type of supervised or semi-supervised machine learning problem with the goal to automatically construct a ranking model from training data.”); and
send a signal to facilitate delivery of the item of targeted content to the device and at least one of the target website or an identified user associated with the target audience based on the predicted likelihood of conversion for the at least one of the target audience or the target website (i.e. facilitate delivery of targeted content based on analysis of the query vector analysis which indicates the predicted likelihood of a conversion occurring) (Grbovic: ¶ [0118] “Advertising may include sponsored search advertising, non-sponsored search advertising, guaranteed and nonguaranteed delivery advertising, ad networks/exchanges, ad targeting, ad serving, and/or ad analytics. Various monetization techniques or models may be used in connection with sponsored search advertising, including advertising associated with user search queries, or non-sponsored search advertising, including graphical or display advertising. In an auction-type online advertising marketplace, advertisers may bid in connection with placement of advertisements, although other factors may also be included in determining advertisement selection or ranking. Bids may be associated with amounts advertisers pay for certain specified occurrences, such as for placed or clicked-on advertisements, for example.” Furthermore, as cited in ¶ [0094] “Additionally, in some embodiments, the web search activity may be weighted based on recency or distance in time from a particular ad click. Traditional skip-gram modeling treats neighboring words as positive when training models and random words as negative. The skip-gram model, however, may be further modified to be account for the issues encountered when analyzing web search data. In particular, instead of treating randomly appearing words as a negative training on the model, the modeling techniques can be adapted to weight more heavily the activity that is closest to an ad click as that activity is most likely to be correlated to the resulting click.” Furthermore, as cited in ¶ [0121] “One approach to presenting targeted advertisements includes employing demographic characteristics (e.g., age, income, sex, occupation, etc.) for predicting user behavior, such as by group. Advertisements may be presented to users in a targeted audience based at least in part upon predicted user behavior(s). Another approach includes profile-type ad targeting. In this approach, user profiles specific to a user may be generated to model user behavior, for example, by tracking a user's path through a web site or network of sites, and compiling a profile based at least in part on pages or advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users.”).
Grbovic does not explicitly disclose identify at least one of a target audience from the plurality of audiences or a target website from the plurality of websites based on a distance, in the word embedding, between the search query location and a location of the at least one of the target audience or the target website.
However, Jamali further discloses identify at least one of a target audience from the plurality of audiences or a target website from the plurality of websites based on a distance, in the word embedding, between the search query location and a location of the at least one of the target audience or the target website (i.e. identify final content or final users /target audience based on a distance or difference or dot product within the modelling or embedding between the keyword and user) (Jamali: ¶¶ [0060] [0061] “A result of inputting initial user-level embedding 324 into neural network 334 is a final user-level embedding 344. Although final user-level embedding 344 is depicted as being a vector of size five, an actual output embedding may be a vector of a larger size, such as ten…At block 280, for each identified content item, an operation on the output of the first neural network (i.e., final content item-level embedding) and the output of the second neural network (i.e., the final user-level embedding) is performed to generate a result. The operation may be a dot product, difference, or summation. The more similar the outputs of the respective neural networks, the more likely the corresponding user will select ( or otherwise interact with) the corresponding content item. Any similarity can be used as a signal for down-stream interaction ( e.g., selection). As a specific example, 1/(1-e'-(L1 *L2)) is computed, where L1 is the final content item-level embedding produced by the first neural network, L2 is the final user-level embedding produced by the second neural network, and '*' is a dot product operation. The result of this computation reflects a probability that the user corresponding to L2 will select the content item corresponding to L1.”).
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to add Jamali’s identify at least one of a target audience from the plurality of audiences or a target website from the plurality of websites based on a distance, in the word embedding, between the search query location and a location of the at least one of the target audience or the target website to Grbovic’s identify at least one of […] a target website from the plurality of websites based on a distance, in the word embedding, between the search query location and a location of the at least one of the target audience or the target website. One of ordinary skill in the art would have been motivated to do so because “automatically learning latent representations of different attribute values and generating embeddings therefrom improves the accuracy of predicted entity selection rates, resulting in identifying more relevant content items for presentation to requesting entities.” (Jamali: ¶ [0011]).
Grbovic and Jamali do not explicitly disclose receive, from a device that has cookie-based tracking disabled, a search query associated with an item of targeted content.
However, Li further discloses receive, from a device that has cookie-based tracking disabled, a search query associated with an item of targeted content (i.e. user device has disabled cookie-based tracking, wherein the user device making the query is not identified because the device is associated with a user behavior type) (Li: ¶¶ [0025]-[0027] “This problem may also exist if a user disables cookies on their web browser, and such problems are especially prevalent in users of mobile devices that utilize web browsers which don't support third-party cookie capabilities or are set up to disable cookies by default…In their most basic form, the system and methods of the present description seek to generate and utilize a multi-indexed data structure representing user behavior, also referred to herein as a poly-indexed user behavior data cube or hypercube, to associate consumer behaviors with retrievable data in a cookie-disabled system, such as from cookie-disabled mobile browsers. In certain embodiments, the system stores user behavior data relating to prior offline and online activities of a user. The user behavior data is processed to generate indices defining relationships between attribute data and user behaviors or events, such as interactions with monitored web content. The generated indices can then be assembled and associated into an aggregated data structure defining relationships between the various indices. In this way, the aggregated data structure can be made up of a series of indices, wherein each of the indices relates to a particular user behavior associated with a given set of user attributes…In one embodiment, the aggregated data structure takes the form of a poly-indexed hypercube. Each index may be accessible, for example, by using a subset of the attribute data as key for the index, such as one or more traditional targeting data or behavioral targeting data, as discussed above. The data structure may be augmented with additional data as further user behavior data is received and stored by the system. In this way, the poly-indexed hypercube can "grow" or expand to include additional indices representing behaviors associated with additional subsets of attribute data. In accordance with further embodiments described herein, performance data may be used to optimize associations between indices and user behavior data. Additionally, performance data may also be used to optimize utilization of the poly-indexed hypercube to select advertisements from an advertisement cache.” Furthermore, as cited in ¶¶ [0040] [0041] “In certain embodiments, it is preferable not to positively identify the user who makes the query using a mobile device, regardless of whether the user is or is not listed in the user behavior database 210. Rather, in such embodiments, the identity of such a user making the query from a mobile device may be maintained in private manner, if desired, through the use of informed behavioral retargeting. For example, while certain embodiments may make use of device IDs, device fingerprint IDs, and the like, a user's identify need not be verified or confirmed, as the system is able to accurately serve retargeted advertisements to that user through the user of the multi-indexed user behavior cube 220, in accordance with various systems and methods as described herein…Once the user behavior database 210 has been populated with data, either initially or as an ongoing updating process, the data is indexed by index lookup module 214. Unlike known retargeting systems, the input data in the user behavior database 210 is not indexed solely by cookie-type identification, but rather indexing may make use of a variety of different attribute data in addition to cookie-type data.”)).
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to add Li’s receive, from a device that has cookie-based tracking disabled, a search query associated with an item of targeted content to Grbovic’s receive a search query associated with an item of targeted content. One of ordinary skill in the art would have been motivated to do so because “it is desirable to provide an alternative system and method for targeting advertising to consumers based on both traditional targeting data and behavioral targeting data, such as their previous online and offline activities.” (Li: ¶ [0025]).
With respect to Claim 3:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 2, wherein the training data includes the behavioral data on which the location of each audience from the plurality of audiences is based (i.e. define training data to be query words in vector space that lead user through navigation to click or other interactions such as purchases) (Grbovic: ¶ [0023] “In particular, certain embodiments are directed systems and methods for generating data-driven keyword clusters using distributed query word representations formed from novel techniques for analyzing and processing historical web search activity, including, by way of illustration, historical search queries entered by users, historical advertising campaigns, recorded ad clicks or interactions, ad impressions, and resulting ad conversions, for example. Keyword cluster sets or keyword lists for a specific advertiser or campaign type may be generated by learning distributed representations of user queries that are most likely to lead to ad clicks and conversions. In some embodiments, the distributed representation may be generated by applying a directed approach to learning distributed representations that focuses on or weights the data to emphasize actions immediately preceding an ad click. For example, by using deep learning technologies, the circuitry components of the present system generate distributed representations of query words in vector space using the search engine data, such that similar words in context of web search (i.e., those that are most likely to lead to ad clicks) can be found in a cluster K of the nearest neighbors of an adword or keyword category.”).
With respect to Claim 4:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 3, wherein the behavioral data includes aggregated behavioral data (Grbovic: ¶ [0083] “As described in connection with FIG. 4, the web search data is typically aggregated on a per-user basis in order to form profiles for targeting. For example, raw activity logs of search queries with timestamps may be stored for every user. The activity for each user is recorded as one record in the activity logs. The system may retrieve all web search data for a recent period of time, such as for the past six months, and examine the data on a per-user basis to determine keyword relevancy to the particular user. In this way, the system can ultimately generated targeted keyword lists for a particular user, or set of users determined to be similar based on known profile traits, in order to provide search retargeting rules that target the particular user or set of users having similar traits.”).
With respect to Claim 18:
All limitations as recited have been analyzed and rejected to claim 4. Claim 18 does not teach or define any new limitations beyond claim 4. Therefore it is rejected under the same rationale.
With respect to Claim 5:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 2, the code further comprising code to cause the processor to define the plurality of audiences based on the behavioral data, each audience from the plurality of audiences including at least one user from the plurality of users (i.e. audience segments are defined according to behavioral data such as user interest and intentions) (Grbovic: ¶ [0038] “Advertising may be beneficial to users, advertisers or web portals if displayed advertisements are relevant to audience segments, segment combinations, or at least parts of campaigns. Thus, a variety of techniques have been developed to determine corresponding audience segments or to subsequently target relevant advertising to audience members of such segments. For example user interests, user intentions, and targeting data related to segments or campaigns may be may be logged in data logs and such logs may be communicated to the analytics server 118 for processing.”).
With respect to Claim 6:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 2, the code further comprising code to cause the processor to define at least one audience from the plurality of audiences based on the behavioral data, the at least one audience defined based, at least in part, on an indication in the behavioral data that an audience member of the at least one audience visited at least one website (i.e. audience segments are defined based on user behavior such as user’s visitation or user’s path through website) (Grbovic: ¶¶ [0039] [0040] “One approach to presenting targeted advertisements includes employing demographic characteristics (such as age, income, sex, occupation, etc.) for predicting user behavior, such as by group. Advertisements may be presented to users in a targeted audience based, at least in part, upon predicted user behavior. The aforementioned targeting data, such as demographic data and psychographic data, may be logged in data logs and such logs may be communicated to the analytics server 118 for processing. Once processed into corresponding analytics data, the analytics server 118 can provide analyzed feedback for affecting future serving of content.…Another approach includes profile-type ad targeting. In this approach, user or group profiles specific to a respective user or group may be generated to model user behavior, for example, by tracking a user's path through a website or network of sites, and compiling a profile based, at least in part, on ad GUIs, webpages, and advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users. The aforementioned profile-type targeting data may be logged in data logs and such logs may be communicated to the analytics server 118 for processing.”).
With respect to Claim 7:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 2, the code further comprising code to cause the processor to: access the behavioral data (i.e. access historical ad clicks which is sessionalized according to user segments or plurality of audiences) (Grbovic: ¶ [0083] “At block 518, the system circuitry retrieves historical web search data related the identified ad categories from the system databases. By way of illustration, the historical web search data may include historical search queries entered by users, historical advertising campaigns, recorded ad clicks or interactions with ad content, ad impressions, and resulting ad conversions, for example…The activity for each user is recorded as one record in the activity logs. The system may retrieve all web search data for a recent period of time, such as for the past six months, and examine the data on a per-user basis to determine keyword relevancy to the particular user. In this way, the system can ultimately generated targeted keyword lists for a particular user, or set of users determined to be similar based on known profile traits, in order to provide search retargeting rules that target the particular user or set of users having similar traits.”); and
the code to cause the processor to determine the location in the word embedding associated with at least one of each website from the plurality of websites or each audience from the plurality of audiences includes code to cause the processor to determine a location in the word embedding for each audience from the plurality of audiences, each audience from the plurality of audiences including at least one user from a plurality of users, the location in the word embedding determined based on behavioral data for users from the plurality of users that are associated with that audience (i.e. determine user’s navigation path in model/embedding for each user profile type, wherein the navigation path is determined based on clicks/purchases or behavioral data for users with the same user profile type) (Grbovic: ¶ [0040] “Another approach includes profile-type ad targeting. In this approach, user or group profiles specific to a respective user or group may be generated to model user behavior, for example, by tracking a user's path through a website or network of sites, and compiling a profile based, at least in part, on ad GUIs, webpages, and advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users. The aforementioned profile-type targeting data may be logged in data logs and such logs may be communicated to the analytics server 118 for processing. Once processed into corresponding analytics data, the analytics server 118 can provide analyzed feedback for affecting future serving of content.” Furthermore, as cited in ¶ [0121] “In this approach, user profiles specific to a user may be generated to model user behavior, for example, by tracking a user's path through a web site or network of sites, and compiling a profile based at least in part on pages or advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users.”).
With respect to Claim 19:
All limitations as recited have been analyzed and rejected to claim 7. Claim 19 does not teach or define any new limitations beyond claim 7. Therefore it is rejected under the same rationale.
With respect to Claim 8:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 2, the code further comprising code to cause the processor to define the word embedding (Grbovic: ¶ [0079] “At block 414, the system circuitry applies one or more modified linguistic modeling or statistical natural language processing techniques, such as a modified skip-gram model in some embodiments, to the results of the pre-processing in order to identify distributed query word representations in the historical web search data.”).
With respect to Claim 16:
All limitations as recited have been analyzed and rejected to claim 8. Claim 16 does not teach or define any new limitations beyond claim 8. Therefore it is rejected under the same rationale.
With respect to Claim 9:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 2, the code further includes code to cause the processor to: rank at least a subset of audiences from the plurality of audiences based on a distance between the search query location in the word embedding and the position of each audience from the subset of audiences in the word embedding, the target audience being a shortest distance from the search query in the word embedding (i.e. ranking based on distance or weight between search query and position in the embedding) (Grbovic: ¶ [0114] “Search results located during a search of an index performed in response to a search query submission may typically be ranked. An index may include entries with an index entry assigned a value referred to as a weight. A search query may comprise search query terms, wherein a query term may correspond to an index entry. In an embodiment, search results may be ranked by scoring located files or records, for example, such as in accordance with number of times a query term occurs weighed in accordance with a weight assigned to an index entry corresponding to the query term. Other aspects may also affect ranking, such as, for example, proximity of query terms within a located record or file, or semantic usage, for example. A score and an identifier for a located record or file, for example, may be stored in a respective entry of a ranking list. A list of search results may be ranked in accordance with scores, which may, for example, be provided in response to a search query. In some embodiments, machine-learned ranking (MLR) models are used to rank search results. MLR is a type of supervised or semi-supervised machine learning problem with the goal to automatically construct a ranking model from training data.”).
With respect to Claim 17:
All limitations as recited have been analyzed and rejected to claim 9. Claim 17 does not teach or define any new limitations beyond claim 9. Therefore it is rejected under the same rationale.
With respect to Claim 10:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 2, wherein the search query is associated with at least one of a summary characteristic associated with the target audience for the item of targeted content or a descriptive title associated with at the target audience for the item of targeted content (i.e. search query is associated to a descriptive title such as categories of products associated with the audience segment interests or categories of targeted audience types such as known profile traits) (Grbovic: ¶¶ [0074] [0075] “FIG. 4 illustrates exemplary operations that may be performed, according to one embodiment, by the circuitry of a search retargeting server in an exemplary system in order to generate distributed query representations to be used for search retargeting. At block 402, the advertiser accesses the system interface of the search retargeting server (or ad server) and creates an advertisement campaign. Alternatively, or in addition, the advertiser may submit an existing campaign…The categories are often based on market research and include standard sets of keywords that advertisers use for campaigns. For example, the advertiser may have a set of keywords that uses for all "travel" related ads. In other examples, the categories may include keywords associated with competing brands and manufacturers that the advertiser wishes to use to retarget. Beginning with the step at block 404, the system may start with the determined ad category, optionally including any generic list of keywords related to the category provided by the advertiser, and produce a more comprehensive, exhaustive, and highly targeted list of ad keywords.” Furthermore, as cited in ¶ [0083] “The activity for each user is recorded as one record in the activity logs. The system may retrieve all web search data for a recent period of time, such as for the past six months, and examine the data on a per-user basis to determine keyword relevancy to the particular user. In this way, the system can ultimately generated targeted keyword lists for a particular user, or set of users determined to be similar based on known profile traits, in order to provide search retargeting rules that target the particular user or set of users having similar traits.”).
With respect to Claim 15:
All limitations as recited have been analyzed and rejected to claim 10. Claim 15 does not teach or define any new limitations beyond claim 10. Therefore it is rejected under the same rationale.
With respect to Claim 11:
Grbovic teaches:
A non-transitory, processor-readable medium storing code representing instructions to be executed by a processor, the code comprising code to cause the processor to (Grbovic: ¶ [0045]):
access a word embedding, the word embedding defined based on a natural language corpus (i.e. access word representation defined based on natural language corpus, the word representation sessionalized for user segment or user profile type including a first and second plurality of audiences or user segments, and the word representation includes keywords/query terms that lead user throughout a path within a webpage that may lead to a conversion) (Grbovic: ¶ [0079] “At block 414, the system circuitry applies one or more modified linguistic modeling or statistical natural language processing techniques, such as a modified skip-gram model in some embodiments, to the results of the pre-processing in order to identify distributed query word representations in the historical web search data. In some embodiments, the distributed query word representations consist of associations between search query terms and ad clicks to the actions of a user. For example, the distributed query word representations may represent a likelihood that a user will perform a given action ( e.g., click on a displayed ad related to a particular category) after the user enters a search query containing a particular keyword. Traditional natural language processing techniques may typically involve one or more algorithms performed on an article, set of articles, or similar body of text that are input into to the algorithm and treated as "documents." Each "word" in the document is then analyzed to determine the statistical relevance.” Furthermore, as cited in ¶ [0040] “Another approach includes profile-type ad targeting. In this approach, user or group profiles specific to a respective user or group may be generated to model user behavior, for example, by tracking a user's path through a website or network of sites, and compiling a profile based, at least in part, on ad GUIs, webpages, and advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users. The aforementioned profile-type targeting data may be logged in data logs and such logs may be communicated to the analytics server 118 for processing. Once processed into corresponding analytics data, the analytics server 118 can provide analyzed feedback for affecting future serving of content.”);
access behavioral data associated with a plurality of users (i.e. access historical ad clicks which is sessionalized according to user segments or plurality of audiences) (Grbovic: ¶ [0083] “At block 518, the system circuitry retrieves historical web search data related the identified ad categories from the system databases. By way of illustration, the historical web search data may include historical search queries entered by users, historical advertising campaigns, recorded ad clicks or interactions with ad content, ad impressions, and resulting ad conversions, for example…The activity for each user is recorded as one record in the activity logs. The system may retrieve all web search data for a recent period of time, such as for the past six months, and examine the data on a per-user basis to determine keyword relevancy to the particular user. In this way, the system can ultimately generated targeted keyword lists for a particular user, or set of users determined to be similar based on known profile traits, in order to provide search retargeting rules that target the particular user or set of users having similar traits.”);
determine a location in the word embedding for each audience from the plurality of audiences, each audience from the plurality of audiences including at least one user from the plurality of users, the location in the word embedding determined based on behavioral data for users from the plurality of users that are associated with that audience (i.e. determine user’s navigation path in model/embedding for each user profile type, wherein the navigation path is determined based on clicks/purchases or behavioral data for users with the same user profile type) (Grbovic: ¶ [0040] “Another approach includes profile-type ad targeting. In this approach, user or group profiles specific to a respective user or group may be generated to model user behavior, for example, by tracking a user's path through a website or network of sites, and compiling a profile based, at least in part, on ad GUIs, webpages, and advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users. The aforementioned profile-type targeting data may be logged in data logs and such logs may be communicated to the analytics server 118 for processing. Once processed into corresponding analytics data, the analytics server 118 can provide analyzed feedback for affecting future serving of content.” Furthermore, as cited in ¶ [0121] “In this approach, user profiles specific to a user may be generated to model user behavior, for example, by tracking a user's path through a web site or network of sites, and compiling a profile based at least in part on pages or advertisements ultimately delivered. A correlation may be identified, such as for user purchases, for example. An identified correlation may be used to target potential purchasers by targeting content or advertisements to particular users.”);
receive a search query associated with an item of targeted content (i.e. receive search query) (Grbovic: ¶ [0064] “FIG. 2 illustrates a block diagram of circuitry components of a sponsored verb generator according to some embodiments. Keyword vector generator 200 may be communicatively coupled search retargeting framework server 116 and may include retargeting circuitry 202, modeling circuitry 204, training circuitry 206, and/or display logic circuitry 208 components. Search retargeting framework server 116 may receive a search query to from a user device and determine one or more search suggestions, sponsored or non-sponsored search results, advertisements, or other related ad content to display to the user.”);
determine a search query location in the word embedding based on the search query (i.e. determine search query location in vector embedding based on search query) (Grbovic: ¶ [0064] “For each search query entered by the user, search retargeting framework server 116 may seek to identify opportunities for monetization, including by using search retargeting rules for keyword lists that have been generated by keyword vector generator 200 using directed distributed query word representations. Search retargeting framework server 116 will communicate the requests containing search query words to keyword vector generator 200. The request will be received by keyword vector generator 200 and retargeting circuitry 202 will determine one or retargeting rules to be used in selecting an advertisement or sponsored content for display to the user request.” Furthermore, as cited in ¶¶ [0079] [0080] “For each search query entered by the user, search retargeting framework server 116 may seek to identify opportunities for monetization, including by using search retargeting rules for keyword lists that have been generated by keyword vector generator 200 using directed distributed query word representations. Search retargeting framework server 116 will communicate the requests containing search query words to keyword vector generator 200. The request will be received by keyword vector generator 200 and retargeting circuitry 202 will determine one or retargeting rules to be used in selecting an advertisement or sponsored content for display to the user request…In some embodiments, the result of steps 414 and 416 is in the form of a vector representing the keyword distributions as related to the input category or advertiser. In these embodiments, as well as others, the keywords that are most closely related to the input advertiser name or ad category are represented in the vector as being nearest to the advertiser name or ad category. In this way, the set of the most closely related keywords in the vector representations can be selected as having the highest likelihood that they are indicative or predictive of an ad click. In addition, in some embodiments, the system circuitry may generate a set of retargeting rules using the keyword list and the closest K neighbors in the list to be used in conjunction with search retargeting techniques.”); […] and
send a signal to facilitate delivery of the item of targeted content, based on the distance, to an identified user associated with the target audience (i.e. facilitate deliver of targeted content based on analysis of the query vector analysis) (Grbovic: ¶ [0118] “Advertising may include sponsored search advertising, non-sponsored search advertising, guaranteed and nonguaranteed delivery advertising, ad networks/exchanges, ad targeting, ad serving, and/or ad analytics. Various monetization techniques or models may be used in connection with sponsored search advertising, including advertising associated with user search queries, or non-sponsored search advertising, including graphical or display advertising. In an auction-type online advertising marketplace, advertisers may bid in connection with placement of advertisements, although other factors may also be included in determining advertisement selection or ranking. Bids may be associated with amounts advertisers pay for certain specified occurrences, such as for placed or clicked-on advertisements, for example.” Furthermore, as cited in ¶ [0094] “Additionally, in some embodiments, the web search activity may be weighted based on recency or distance in time from a particular ad click. Traditional skip-gram modeling treats neighboring words as positive when training models and random words as negative. The skip-gram model, however, may be further modified to be account for the issues encountered when analyzing web search data. In particular, instead of treating randomly appearing words as a negative training on the model, the modeling techniques can be adapted to weight more heavily the activity that is closest to an ad click as that activity is most likely to be correlated to the resulting click.”).
Grbovic does not explicitly disclose identify a target audience from the plurality of audiences based on a distance between the location in the word embedding associated with the target audience and the search query location.
However, Jamali further discloses identify a target audience from the plurality of audiences based on a distance between the location in the word embedding associated with the target audience and the search query location (i.e. identify final users or target audience based on a distance or difference or dot product within the modelling or embedding between the keyword and user) (Jamali: ¶¶ [0060] [0061] “A result of inputting initial user-level embedding 324 into neural network 334 is a final user-level embedding 344. Although final user-level embedding 344 is depicted as being a vector of size five, an actual output embedding may be a vector of a larger size, such as ten…At block 280, for each identified content item, an operation on the output of the first neural network (i.e., final content item-level embedding) and the output of the second neural network (i.e., the final user-level embedding) is performed to generate a result. The operation may be a dot product, difference, or summation. The more similar the outputs of the respective neural networks, the more likely the corresponding user will select ( or otherwise interact with) the corresponding content item. Any similarity can be used as a signal for down-stream interaction ( e.g., selection). As a specific example, 1/(1-e'-(L1 *L2)) is computed, where L1 is the final content item-level embedding produced by the first neural network, L2 is the final user-level embedding produced by the second neural network, and '*' is a dot product operation. The result of this computation reflects a probability that the user corresponding to L2 will select the content item corresponding to L1.”).
Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to add Jamali’s identify a target audience from the plurality of audiences based on a distance between the location in the word embedding associated with the target audience and the search query location to Grbovic’s • send a signal to facilitate delivery of the item of targeted content, based on the distance, to an identified user associated with the target audience. One of ordinary skill in the art would have been motivated to do so because “automatically learning latent representations of different attribute values and generating embeddings therefrom improves the accuracy of predicted entity selection rates, resulting in identifying more relevant content items for presentation to requesting entities.” (Jamali: ¶ [0011]).
With respect to Claim 12:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 11, wherein the code further comprises code to cause the one or more processors to train a machine learning model to predict a likelihood of conversion of the targeted content, given a location in the word embedding (i.e. training deep learning model using word representation and conversion event data in order to predict a likelihood that a content will receive a click or conversion) (Grbovic: ¶ [0023] “In particular, certain embodiments are directed systems and methods for generating data-driven keyword clusters using distributed query word representations formed from novel techniques for analyzing and processing historical web search activity, including, by way of illustration, historical search queries entered by users, historical advertising campaigns, recorded ad clicks or interactions, ad impressions, and resulting ad conversions, for example. Keyword cluster sets or keyword lists for a specific advertiser or campaign type may be generated by learning distributed representations of user queries that are most likely to lead to ad clicks and conversions. In some embodiments, the distributed representation may be generated by applying a directed approach to learning distributed representations that focuses on or weights the data to emphasize actions immediately preceding an ad click. For example, by using deep learning technologies, the circuitry components of the present system generate distributed representations of query words in vector space using the search engine data, such that similar words in context of web search (i.e., those that are most likely to lead to ad clicks) can be found in a cluster K of the nearest neighbors of an adword or keyword category.” Furthermore, as cited in ¶ [0066] “In some embodiments, modeling circuitry 204 uses computational linguistic analysis techniques that utilize aspects of skip-gram modeling to process web search activity. Instead of processing word and documents, the modified modeling program processes historical web search activity, treating ad clicks and search queries in a manner akin to how one may treat words of a document in linguistic analysis. The modeling techniques are further adapted to consider time-related data associated with the web search activity, such that the algorithm is time-sensitive. In this way, the system circuitry, including modeling circuitry 204, can generate vector representations of keywords that are statistically indicative of the correlation between ad clicks, search query terms, and targeting keywords. In other words, modeling circuitry 204 generates vector representations of the likelihood that a keyword is related to a category of an advertisement that the keyword is likely to lead to an ad click.”).
With respect to Claim 13:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 11, wherein the behavioral data for a user from the plurality of users is generated at a user device belonging to that user (i.e. impressions or interactions with content are logged via client device) (Grbovic: ¶ [0034] “The search engine server 106, the search retargeting framework server 116, the sponsored search server 117, or any combination thereof, may receive user interaction information, that can include search queries, from an audience device, and send corresponding information to the ad server 108 and/or the content server 112, and the ad server 108 and/or the content server 112 may serve corresponding ads and/or search results, but with more in-depth details or accompanying GUIs and subGUIs for interacting with subject matter associated with ads or other sponsored content. The information inputted and/or outputted by these devices may be logged in data logs and communicated to the analytics server 118 over the network 120 for processing by the analytics circuitry. The analytics server 118 and related circuitry can provide analyzed feedback for affecting future serving of content.”).
With respect to Claim 14:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 11, wherein: the code to cause the processor to receive the search query includes code to cause the one or more processors to receive the item of targeted content (i.e. receive search query to display targeted content to the user) (Grbovic: ¶ [0064] “FIG. 2 illustrates a block diagram of circuitry components of a sponsored verb generator according to some embodiments. Keyword vector generator 200 may be communicatively coupled search retargeting framework server 116 and may include retargeting circuitry 202, modeling circuitry 204, training circuitry 206, and/or display logic circuitry 208 components. Search retargeting framework server 116 may receive a search query to from a user device and determine one or more search suggestions, sponsored or non-sponsored search results, advertisements, or other related ad content to display to the user.”); and
the code to cause the processor to determine the search query location includes code to cause the one or more processors to generate the search query location based on an analysis of the item of targeted content (i.e. determine search query location in vector embedding based on analysis of targeted content) (Grbovic: ¶ [0064] “For each search query entered by the user, search retargeting framework server 116 may seek to identify opportunities for monetization, including by using search retargeting rules for keyword lists that have been generated by keyword vector generator 200 using directed distributed query word representations. Search retargeting framework server 116 will communicate the requests containing search query words to keyword vector generator 200. The request will be received by keyword vector generator 200 and retargeting circuitry 202 will determine one or retargeting rules to be used in selecting an advertisement or sponsored content for display to the user request.” Furthermore, as cited in ¶¶ [0079] [0080] “For each search query entered by the user, search retargeting framework server 116 may seek to identify opportunities for monetization, including by using search retargeting rules for keyword lists that have been generated by keyword vector generator 200 using directed distributed query word representations. Search retargeting framework server 116 will communicate the requests containing search query words to keyword vector generator 200. The request will be received by keyword vector generator 200 and retargeting circuitry 202 will determine one or retargeting rules to be used in selecting an advertisement or sponsored content for display to the user request…In some embodiments, the result of steps 414 and 416 is in the form of a vector representing the keyword distributions as related to the input category or advertiser. In these embodiments, as well as others, the keywords that are most closely related to the input advertiser name or ad category are represented in the vector as being nearest to the advertiser name or ad category. In this way, the set of the most closely related keywords in the vector representations can be selected as having the highest likelihood that they are indicative or predictive of an ad click. In addition, in some embodiments, the system circuitry may generate a set of retargeting rules using the keyword list and the closest K neighbors in the list to be used in conjunction with search retargeting techniques.”).
With respect to Claim 20:
Grbovic teaches:
The non-transitory, processor-readable medium of claim 11, wherein the distance between the location in the word embedding associated with the target audience and the search query location represents a quantitative semantic similarity between the search query and behavioral data associated with the target audience (i.e. distance between query and behavior associated with audience such as click activity, wherein the distance represents semantic similarity) (Grbovic: ¶ [0095] “At block 546, the system circuitry trains the modified linguistic model, which may be a modified skip-gram model, based on the search query terms present in the web search data only. These steps may be applied to the search query terms only as they account for a major source of the semantic issues present in the web search data that consist of query search terms that are not necessarily present with ad clicks. In particular, at block 548, the system circuitry processes the search query terms to identify common spelling mistakes. For example, in one embodiment, the Damerau-Levenshtein distance between two words may be used to identify misspellings. The Damerau-Levenshtein distance between two words is the count of operations needed to transform the first word into the second word, where operations include insertion, deletion, or substitution of a single character, as well as transpositions. For most words in the natural language misspellings are typically at distance 2 or less. Therefore, among the top 200 neighbors of a particular search query term the system is able to find those that are at distance 2 or less, for example, and treat them as misspellings of the same term for the purposes of processing the web search data. Similarly, at block 550, the system circuitry processes the results to identify plural forms of the same search query terms within each session. At block 552, once the misspellings and plurals are identified, the system can replace all wrong spellings and plurals with correct spellings or same form of the term and retrains the model using these changes to the list of search query terms in the list of ad keywords generated at block 542.”).
Response to Arguments
Applicant’s arguments see page 7 of the Remarks disclosed, filed on 03/27/2026, with respect to the nonstatutory double patenting rejection(s) of claim(s) 2-20 have been considered but are not persuasive. The Applicant has stated “Claims 2-20 stand rejected on the ground of nonstatutory double patenting as allegedly being unpatentable over claims 1-25 of U.S. Patent No. 11,068,935. Claims 2-20 stand rejected on the ground of nonstatutory double patenting as allegedly being unpatentable over claims 1-21 of U.S. Patent No. 11,699,109. Claims 2-20 stand rejected on the ground of nonstatutory double patenting as allegedly being unpatentable over claims 1-29 of U.S. Patent No. 12,045,703. Each of these rejections will be traverse when the claims are otherwise in condition for allowance.” The Examiner respectfully disagrees. Therefore, the rejection(s) of claim(s) 2-20 under the nonstatutory double patenting rejection is maintained above.
Applicant’s arguments see pages 11-14 of the Remarks disclosed, filed on 07/22/2025, with respect to the 35 U.S.C. § 101 rejection(s) of claim(s) 2-20 have been considered but are not persuasive. The Applicant asserts “Without acceding to the rejection of claim 2, in the interest of compact prosecution, claim 2 has been amended to recite, among other things, "determine a location in the word embedding associated with... each audience from a plurality of audiences associated with devices that have behavioral-based tracking enabled," "receive, from a device that has cookie-based tracking disabled, a search query associated with an item of targeted content," and "send a signal to facilitate delivery of the item of targeted content to the device" (emphasis added). Applicant contends that claim 2 as amended is patent eligible for at least the following two reasons. First, claim 2 as amended is not directed to an abstract idea. For example, claim 2 is not a mental process because the human mind cannot practically enable, disable, or perform behavioral-based tracking. 1 As another example, claim 2 is not a certain method of organizing human activity because disabling or performing cookie-based tracking is fundamentally different from recognized methods of organizing human activity.²… Here, Applicant's specification discloses-and claim 2 reflects-a technical solution to a technical problem. For example, Applicant's paragraph [0007] discloses a technical problem: "individual users have been classified using an identifier shared by the user's device, such as a cookie identifier... [r]ecently, however, there has been a renewed effort by browser developers to enable private-browsing technologies that prevent content providers from identifying individuals." In this context of being unable to identify individuals due to cookies being inaccessible, however, Applicant discloses a technical solution to nonetheless send targeted content to a device that has behavioral-based tracking disabled, such as using a word embedding, determining various locations in the word embedding, defining training data, and training a machine learning model. Claim 2 further reflects the technical problem, reciting "a device that has cookie-based tracking disabled." That said, despite a technical problem that a device has cookie-based tracking disabled, claim 2 also recites a technical solution that includes steps like "[accessing] a word embedding," [determining] a location in the word embedding," [defining] training data," and "[training] a machine learning model" that ultimately allows "[sending] a signal to facilitate delivery of the item of targeted content to the device." Said similarly, claim 2 recites steps that result in sending targeted content to a device that has cookie-based tracking disabled.” The Examiner respectfully disagrees. The behavioral tracking or cookie-based tracking is recited at such a high level in the claims (e.g. “audiences associated with devices that have behavioral-based tracking enabled” and “from a device that has cookie-based tracking disabled”) that the behavioral tracking being disabled/enabled merely describes computing environment (i.e. device) that is being used to implement the abstract idea. Furthermore, Claims 2 and 11 recite additional limitation “a non-transitory, processor-readable medium storing code representing instructions to be executed by a processor; associated with devices that have behavioral-based tracking enabled; and from a device that has cookie-based tracking disabled”. Claim 2 recites additional limitations including “define training data based on determining a location in the word embedding for each website visited that is associated with a conversion action taken; train a machine learning model using the training data to predict a likelihood of conversion based on a location in the word embedding; and apply the machine learning model to the location in the word embedding of the at least one of the target audience or the target website to predict a likelihood of conversion for the at least one of the target audience or the target website”. Claims 2 and 11 reciting additional limitations including “a non-transitory, processor-readable medium storing code representing instructions to be executed by a processor; associated with devices that have behavioral-based tracking enabled; and from a device that has cookie-based tracking disabled” do not integrate the claims into practical application. Merely reciting the processor-readable medium and devices that have behavioral-based tracking enabled/disabled in this manner is seen as adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f). Accordingly, alone, and in combination, these additional elements are seen as using a computer or tool to perform an abstract idea, adding insignificant-extra-solution activity to the judicial exception. They do no more than link the judicial exception to a particular technological environment or field of use, i.e. processor-readable medium and devices, and therefore do not integrate the abstract idea into a practical application. The courts decided that although the additional elements did limit the use of the abstract idea, the court explained that this type of limitation merely confines the use of the abstract idea to a particular technological environment and this fails to add an inventive concept to the claims (See Affinity Labs of Texas v. DirecTV, LLC,). Therefore, the rejection(s) of claim(s) 2-20 under 35 U.S.C. § 101 is maintained above with an updated analysis.
Applicant’s arguments see pages 10-11 of the Remarks disclosed, filed on 03/27/2026, with respect to the 35 U.S.C. § 103 rejection(s) of claim(s) 2-20 over Grbovic in view of Jamali have been considered but are moot because the arguments do not apply to the new ground(s) of rejection is made in view of U.S. Publication 2014/0297394 to Li.
Conclusion
The prior art made of record and not relied upon is considered pertinent to Applicant’s disclosure. The following reference are cited to further show the state of the art:
U.S. Patent 10,878,480 to Hueter for disclosure of a system and method is disclosed for collecting website visitor activity for profiling visitor interests and dynamically modifying the content of the website to better match the visitor's profile. The visitor activity data is collected directly from the visitor's client browser or from the website's own web log information. The collected data consists of the page identifier, page links, and the previous page identifier. Similarly, the modified page content can be sent directly to the client browser or can be sent back to the website server for integration with the other page content. The collected data is stored in a database. Based on the amount of information collected on the visitor and the various items that are presented on the website, the visitors and items are profiled so that a visitor's response to other items can be predicted and recommended to the visitor.
U.S. Publication 2010/0198772 to Silverman for disclosure of Methods, systems, and apparatus, including computer program products, for determining a probability that a traffic conversion of a content item associated with a content source (e.g., website) will occur based on past traffic patterns for that content source. A traffic conversion defines, for example, minimum traffic interactions of one or more associated user sessions with a content source. The minimum traffic interactions can be based on, for example, the duration of the one or more user sessions on the content source, or a quantity of pages associated with the content source navigated in the one or more associated user sessions.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Azam Ansari, whose telephone number is (571) 272-7047. The examiner can normally be reached from Monday to Friday between 8 AM and 4:30 PM.
If any attempt to reach the examiner by telephone is unsuccessful, the examiner's supervisor, Waseem Ashraf, can be reached at (571) 270-3948.
Another resource that is available to applicants is the Patent Application Information Retrieval (PAIR). Information regarding the status of an application can be obtained from the (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAX. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pairdirect.uspto.gov. Should you have questions on access to the Private PAIR system, please feel free to contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Applicants are invited to contact the Office to schedule either an in-person or a telephonic interview to discuss and resolve the issues set forth in this Office Action. Although an interview is not required, the Office believes that an interview can be of use to resolve any issues related to a patent application in an efficient and prompt manner.
/AZAM A ANSARI/
Primary Examiner, Art Unit 3621
May 30, 2026
Prosecution Insights