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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 February 20, 2026, has been entered.
Claims 1, 8, and 21 are amended.
Claim 7 is canceled.
Claims 1, 3-6, 8-11, 14-19, and 21 are pending.
Information Disclosure Statement
The Information Disclosure Statement filed on March 18, 2026, has been considered.
Response to Arguments
Subject Matter Eligibility Declaration
The Examiner has no suggestions to make regarding a subject matter eligibility declaration. As indicated in the rejection, below, the Examiner believes the present claims to be directed to ineligible subject matter. If the Examiner had suggestions to make to put the claims in condition for allowance, the Examiner would provide those suggestions, in accordance with best practices and the interests of compact prosecution.
Rejection under 35 USC §101
The Applicant submits that the present claims are subject matter eligible because the claims are not directed to any of the categories of abstract ideas. See Remarks pp. 15-16. The Examiner respectfully disagrees, and points to the rejection, below, which concludes that the present claims are directed to a method of organizing human activity. Essentially, the present claims recite steps for determining sponsorship exposure that could be implemented mentally or on paper by a human being, but a general purpose computer employing various methods of machine learning is recited for implementation. For example, the claimed metadata tensors and dropout layers are various known machine learning methods that can be combined for image classification. No apparent improvement in machine learning is recited in the claims. Stating that the analysis of video is frame-by-frame merely recognizes the nature of video or “motion pictures” themselves as a series of images.
The Applicant further contends that the image tensor, metadata tensor, and channel-wise concatenation are rooted in computer vision processing, so the claims are subject matter eligible. See Remarks p. 12. In response, the Examiner submits that merely claiming these known methods does not convey subject matter eligibility. The present application does not provide an improvement in any of these known methods of machine learning. Instead, the claims merely recite the use of these known methods for known purposes. As indicated in the rejection, below, the machine learning elements merely provide a technological environment for implementing the abstract idea of measure exposure to sponsorship. The method relies on image classification which could be performed by a human being, but a general purpose computer employing machine learning is recited to classify the images. Contrary to the Applicant’s assertions, a human being could potentially distinguish a doctored or “deep fake” photo from an original.
The Applicant additionally asserts that the claims provide a practical application of any recited abstract idea because the claims provide an advancement in machine learning. See Remarks p. 17. For essentially the same reasons discussed, above, the Examiner disagrees. Again, the Examiner reiterates that the present claims merely recite the idea of machine learning with various known methods to classify images. Merely listing various forms of machine learning that are performed does not provide an improvement in machine learning. The claims rely on the idea of a solution – “the active/passive detection layer being trained to differentiate a logo that is digitally inserted into a frame from a logo that is captured in an original recording . . . “ See exemplary independent claim 1. An important consideration in determining whether a claim improves technology is the extent to which the claim covers a particular solution to a problem or a particular way to achieve a desired outcome, as opposed to merely claiming the idea of a solution or outcome. See MPEP §2105.05(a). The claims do not recite a manner of providing the solution or achieving the outcome; only the outcome is recited.
The rejection for lack of subject matter eligibility is updated and maintained.
35 USC §103 Rejections
Amendments to the claims changed the scope of the claims, necessitating further search and consideration of the prior art. A new search returned the Dutta and Norris references, cited in the rejections, below. The Applicant’s arguments with respect to the previously presented rejection are moot in light of the newly cited references.
The Applicant additionally contends that Harvey is deficient because Harvey teaches away from real-time processing. See Remarks pp. 25-26. In response, the Examiner points out that computers are understood to operate in real-time. At best, the batch processing taught by Harvey discloses a potential embodiment where processing is delayed. The teaching does not “teach away” from real-time analysis. However, Harvey does explicitly teach real-time access to reports. See ¶[0066].
The rejection of the remaining dependent claims stands or falls with the rejection of the independent claims.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
The Manual of Patent Examining Procedure (MPEP) provides detailed rules for determining subject matter eligibility for claims in §2106. Those rules provide a basis for the analysis and finding of ineligibility that follows.
Claims 1, 3-6, 8-11, 14-19, and 21 are rejected under 35 U.S.C. 101. The claimed invention is directed to non-statutory subject matter because the claimed invention recites a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more. Although claims(s) 1, 3-6, 8-11, 14-19, and 21 are all directed to one of the four statutory categories of invention, the claims are directed to measuring exposure to sponsorship (as evidenced by the preamble of exemplary independent claim 1), an abstract idea. Certain methods of organizing human activity are ineligible abstract ideas, including managing personal behavior or relationships or interactions between people. See MPEP §2106.04(a). The limitations of exemplary claim 1 include: “receive a video source media;” “perform a real-time analysis of the video source media;” “generating one or more brand locations and brand sizes within the videos source media;” “detecting one or more assets;” “receiving . . . a structured metadata vector;” “determine sponsorship exposure metrics;” “output an analyzed video source media;” and “provide the sponsorship metrics at least to a sponsor.” The steps are all steps for managing personal behavior related to the abstract idea of measuring exposure to sponsorship that, when considered alone and in combination, are part of the abstract idea of measuring exposure to sponsorship. Several limitations involve the generation of an algebraic tensor (see exemplary independent claim 1; “generating a base output tensor;” “generating a base meta model output feature tensor;” “form a concatenated single feature tensor), but these steps are highly generalized steps for performing mathematical computations. Mathematical concepts are ineligible abstract ideas. See MPEP §2106.04(a). The dependent claims further recite steps for managing personal behavior that are part of the abstract idea of measuring exposure to sponsorship. These claim elements, when considered alone and in combination, are considered to be abstract ideas because they are directed to a method of organizing human activity which includes measuring the effectiveness of advertising.
Under step 2A of the subject matter eligibility analysis, a claim that recites a judicial exception must be evaluated to determine whether the claim provides a practical application of the judicial exception. Additional elements of the independent claims amount to generic computer hardware that does not provide a practical application (a processor and memory in independent claim 1; and a computing device with a processor and memory in independent claim 8). See MPEP §2106.04(d)[I]. The claims do not recite an improvement to another technology or technical field, nor do they recite an improvement to the functioning of the computer itself. See MPEP §2106.05(a). The independent claims do recite the use of a neural network, but the abstract idea of measuring exposure to sponsorship is generally linked to a neural network environment for implementation. A generic neural network structure with various models representing layers of the network is linked to an output provided to a sponsor. Therefore, the neural network merely amounts to a technological environment for implementing the abstract idea of measuring exposure to sponsorship. The technological environment does not provide a practical application or significantly more than an abstract idea. See MPEP §2106.05(h). The claims require no more than a generic computer (a processor and memory in independent claim 1; and a computing device with a processor and memory in independent claim 8) to implement the abstract idea, which does not amount to significantly more than an abstract idea. See MPEP §2106.05(f). Because the claims only recite use of a generic computer, they do not apply the judicial exception with a particular machine. See MPEP §2106.05(b). For these reasons, the claims do not provide a practical application of the abstract idea, nor do they amount to significantly more than an abstract idea under step 2B of the subject matter eligibility analysis. Using a generic computer to implement an abstract idea does not provide an inventive concept. Therefore, the claims recite ineligible subject matter under 35 USC §101.
Furthermore: an element found to amount to insignificant extra-solution activity in step 2A must be evaluated in step 2B to determine whether the step amounts to more than what is well-understood, routine, and conventional. The use of dropout layers is well-understood, routine, and conventional; as evidenced by ¶[0186] of US 20200075148 A1 to Nguyen. Dropout layers are a known method to reduce overfitting. Therefore, the use of dropout layers does not provide a practical application or significantly more than the recited abstract idea. The claims are directed to an abstract idea without significantly more.
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) 1 and 3-6 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 20110288907 A1 to Harvey et al. (hereinafter ‘HARVEY’) in view of US 20180082123 A1 to Katz et al. (hereinafter ‘KATZ’), US 20200320769 A1 to Chen et al. (hereinafter ‘CHEN’), US 10699167 B1 to Dowdall et al. (hereinafter ‘DOWDALL’), US 20200302223 A1 to Dutta et al. (hereinafter ‘DUTTA’), and US 20220084223 A1 to Norris et al. (hereinafter ‘NORRIS’).
Claim 1 (Currently Amended)
HARVEY discloses a sponsorship exposure metric system comprising: a processor (see ¶[0095]; distributed processing techniques with computational tasks executing on multiple processors) configured to:
receive a video source media (see ¶[0200]-[0201]; video advertising networks) comprising a sponsor message (see abstract and ¶[0071]; an advertising measurement system. A touchpoint includes event sponsorships) and a source entity ID (see ¶[0053], [0073]; a watermark embedding a commercial code or program code);
perform real-time analysis of the video source media (see again abstract and ¶[0071]; an advertising measurement system. A touchpoint includes event sponsorships. See also ¶[0066]; advertisers and others can be offered real-time or batch access to research reports).
HARVEY does not specifically disclose, but KATZ discloses each frame being processed by the processor within an inter-frame interval of the video source media (see ¶[0095] and [0101]-[0103] and Fig. 8; a user defines a bounding box. The computing system may track movement of the object within the box and store estimated bounding box information with respect to subsequent frames. Overlays may include visual bounding boxes around sponsor logos recognized by the above techniques, which may follow movement of the logo from frame to frame, as well as graphical indicators of the media value for that exposure),
the real-time analysis comprising: generating one or more brand locations and brand sizes within the video source media with at least one brand detection neural networks (see ¶[0024]; a neural network classifier for image object recognition), the brand detection neural networks trained to detect a sponsor message (see ¶[0027] and [0054]; the size of the logo within the frame and where the logo appears. Media percentage is measured based on clarity, visibility, size, and placement. See also ¶[0146]; “Feeling Anxious? Grab and CandyBar”);
detecting one or more assets using a sports-specific asset detection model, the sports-specific asset detection model including a plurality of neural networks each trained for a specific sport to detect the assets (see abstract and ¶[0024]; a first set of classification models that are each trained to identify at least one type of scene associated with one or more sports, then features of the media content may be provided to a second set of classification models trained to identify at least one object associated with one or more sports. A neural network as a classifier);
HARVEY does not specifically disclose, but CHEN discloses the plurality of neural networks comprising a detection neural network used by an active/passive detection module (see abstract and ¶[0305]; a deep neural network. Region of interest and object detection), the detection neural network comprising:
a base image model generating a base image output tensor (see ¶[0041] and [0228]; feature vector outputs from image feature extract attributes).
HARVEY does not specifically disclose, but DOWDALL discloses, a base meta model receiving, as input, a structured metadata vector that encodes at least a brand identifier, entity identifier, scene-type identifier, asset identifier, normalized bounding-box coordinates, and image dimensions, generating a base meta model output feature tensor (see col 21, ln 34-49; the metadata may include labeling information 722, such as the labels for the object, probability distributions of the labels, classification score of the labels, etc. For another example, the metadata may include bounding box information 724 for that object, such as locations, distances (such as distance from the autonomous vehicle 110), dimensions, elevations, etc. For still another example, the metadata may include other object information 726, such as number of LIDAR points, colors in the object, lighting on the object, etc. For yet another example, the metadata may include scene information 728, such as the scene number, timestamp of the scene, total number of LIDAR points in the scene, lighting and colors of the scene, etc.).
HARVEY further discloses channel-wise concatenating of the base image output tensor and the base meta model output feature tensor to form a concatenated single feature tensor (see ¶[0041]-[0045]; visual features are vectorized and concatenated, and passed to fully-connected layers for attribute classification).
HARVEY does not specifically disclose, but DUTTA discloses, a merged model comprising three fully connected layers paired with a Rectified Linear Unit, a dropout layer, and a Sigmoid activation to an output (see ¶[0298]; activation functions such as ReLU, sigmoid and dropout can be used in a neural network).
HARVEY further discloses the merged model receiving a concatenated single feature tensor, the concatenated single feature tensor being the base image output tensor and the base meta model output feature tensor merged together, the merged model outputting sponsorship-exposure classification logits (see ¶[0041]-[0045]; visual features are vectorized and concatenated, and passed to fully-connected layers for attribute classification) to reduce overfitting in the detection neural network (see ¶[0245] and [0410]; we did not want to overfit the data, and thus wanted to have a lower number of parameters. Prevent neural network from overfitting. See also ¶[0265]; an additional fully-connected layer for dimensionality reduction through by vector concatenation).
HARVEY does not specifically disclose, but KATZ discloses, based on the analysis, determine sponsorship exposure metrics associated with the sponsor message, the sponsorship exposure metrics including size, position, and blurriness of the sponsorship message (see again ¶[0027] and [0054]; the size of the logo within the frame and where the logo appears. Media percentage is measured based on clarity, visibility, size, and placement).
HARVEY further discloses an active exposure (see ¶[0065]; a clickstream may be defined by digitized records produced by active people-metering technology), or passive exposure (see ¶[0065]; a clickstream may be defined by digitized records produced by passive people-metering technology).
HARVEY does not specifically disclose, but NORRIS discloses, using an active/passive detection layer, the active/passive detection layer being trained to differentiate a logo that is digitally inserted in a frame from a logo that is captured in an original recording of the video source media (see ¶[0004]; signatures in the image can be used to detect fake images, which can be categorized into either passive or active approach. In the passive approach, imaging artifacts due to lens distortion, PRNU, and compression can be used to authenticate an image. An imperceptible watermark can be used in the passive approach).
HARVEY does not specifically disclose, but KATZ discloses, output an analyzed video source media with a box around the asset thereby indicating the location and size of the asset (see ¶[0095] and [0101]-[0103] and Fig. 8; a user defines a bounding box. The computing system may track movement of the object within the box and store estimated bounding box information with respect to subsequent frames. Overlays may include visual bounding boxes around sponsor logos recognized by the above techniques, which may follow movement of the logo from frame to frame, as well as graphical indicators of the media value for that exposure ).
HARVEY further discloses provide the sponsorship exposure metrics at least to a sponsor of the sponsor message (see ¶[0003] and [0007]; advertisers need to know which consumers view their commercials. Provide advertisers and other users with accurate measurements of the efficacy of their media advertising campaigns while promoting and protecting consumer privacy. See also ¶[0071] and [0077]; the advertiser can know which advertising has worked and which has not worked. Matched data can be made accessible in real-time or in batch to interested parties such as advertisers,); and
a memory communicatively coupled to the processor, the memory storing instructions executable by the processor (see ¶[0217]-[0220] and claim 1; a computer memory. The process can be performed using instructions stored on a computer-readable memory).
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors. It would have been obvious to identify the logos using a neural network as taught by KATZ in the system executing the method of HARVEY with the motivation to apply a neural network to predict the presence of logos in images and measure the effectiveness of advertising.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). CHEN discloses predicting garment attributes using a neural network for analysis of online retail and shopping (see abstract and ¶[0001]-[0002]). It would have been obvious for one of ordinary skill in the art at the time of invention to use the neural network as taught by CHEN with the motivation to evaluate video media and prevent overfitting of the neural network.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]) based on data provided in images (see ¶[0195]). DOWDALL discloses perception visualization that includes metadata with labels for an object, bounding box information, dimensions, and scene information. It would have been obvious to include the metadata as taught by DOWDALL in the system executing the method of HARVEY with the motivation to label and provide information on objects in images (see DOWDALL abstract).
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). DUTTA discloses artificial intelligence that uses neural networks with layers comprised of activation functions including ReLU, sigmoid, and dropout. It would have been obvious for one of ordinary skill in the art at the time of invention to include the activation functions as taught by DUTTA in the system executing the method of HARVEY with the motivation to use a neural network to measure the effect of advertising on viewership.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). NORRIS discloses detecting fake images and videos using passive or active approaches. It would have been obvious for one of ordinary skill in the art at the time of invention to include fake image detection as taught by NORRIS in the system executing the method of HARVEY with the motivation to distinguish true positives and false positives in a detection network (see NORRIS ¶[0089] and HARVEY ¶[0069]).
Claim 3 (Original)
The combination of HARVEY, KATZ, CHEN, DOWDALL, DUTTA, and NORRIS discloses the system as set forth in claim 1.
HARVEY further discloses wherein the sponsor message includes one or any combination of the following: a brand name (see ¶[0005], [0074], [0077], and [0091] & claims 1 and 55; a subject brand), a logo, a slogan, a text, a hashtag, a tagged mentioning, a text mentioning, a sports type, a league name, a team name, a social media comment, and a social media description (see ¶[0071 and [0105]]; a touchpoint may be social media. Supplement or revise data with social media records).
Claim 4 (Original)
The combination of HARVEY, KATZ, CHEN, DOWDALL, DUTTA, and NORRIS discloses the system as set forth in claim 1.
HARVEY does not specifically disclose, but KATZ discloses, wherein the analysis includes at least one of the following: an optical character recognition (OCR)-based scene classification of the source media, a description-based scene classification, and a generic machine learning based scene classification (see abstract and ¶[0076]-[0079]; perform optical character recognition to identify a team. Determine which team is the home team. Identify a scene using a classification model. See also ¶[0024]; classifiers use machine learning techniques).
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). KATZ discloses machine learning models and optical character recognition used to classify and identify items of interest. It would have been obvious to identify the logos using a neural network and optical character recognition as taught by KATZ in the system executing the method of HARVEY with the motivation to apply a neural network to predict and identify the presence of logos in images and measure the effectiveness of advertising.
Claim 5 (Original)
The combination of HARVEY, KATZ, CHEN, DOWDALL, DUTTA, and NORRIS discloses the system as set forth in claim 1.
HARVEY further discloses wherein the analysis is based on at least one of the following: a shared brand detection model, a generic brand detection model, and an entity specific brand detection model (see ¶[0005]; measure exposure to a subject product brand.).
Claim 6 (Original)
T The combination of HARVEY, KATZ, CHEN, DOWDALL, DUTTA, and NORRIS discloses the system as set forth in claim 1.
HARVEY further discloses wherein the analysis is based on at least two or more of the following: a generic asset detection model, a shared asset detection model, and one or more sports-specific asset detection models (see ¶[0005]; measure exposure to a subject product brand).
Claim(s) 21 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 20110288907 A1 to HARVEY et al. in view of US 20180082123 A1 to KATZ et al., US 10699167 B1 to DOWDALL et al. US 20200320769 A1 to CHEN et al., and US 20200075148 A1 to NGUYEN et al., US 20200234088 A1 to TAHA et al., and US 20220084223 A1 to NORRIS et al.
Claim 21 (Currently Amended)
HARVEY discloses a computer-implemented method for generating sponsorship-exposure metrics, the method comprising, by at least one processor (see ¶[0095]; distributed processing techniques with computational tasks executing on multiple processors) in communication with a memory (see ¶[0217]-[0220] and claim 1; a computer memory. The process can be performed using instructions stored on a computer-readable memory).
HARVEY does not specifically disclose, but KATZ discloses (a) receiving a digital video stream comprising successive video frames (Fo...F~) (see ¶[0095] and [0101]-[0103] and Fig. 8; a user defines a bounding box. The computing system may track movement of the object within the box and store estimated bounding box information with respect to subsequent frames. Overlays may include visual bounding boxes around sponsor logos recognized by the above techniques, which may follow movement of the logo from frame to frame, as well as graphical indicators of the media value for that exposure),
HARVEY does not specifically disclose, but DOWDALL discloses, and, for each frame, a structured metadata vector that encodes a brand ID, entity ID, scene-type ID, asset ID, normalized bounding-box coordinates, image dimensions, and media-type flags (see col 21, ln 34-49; the metadata may include labeling information 722, such as the labels for the object, probability distributions of the labels, classification score of the labels, etc. For another example, the metadata may include bounding box information 724 for that object, such as locations, distances (such as distance from the autonomous vehicle 110), dimensions, elevations, etc. For still another example, the metadata may include other object information 726, such as number of LIDAR points, colors in the object, lighting on the object, etc. For yet another example, the metadata may include scene information 728, such as the scene number, timestamp of the scene, total number of LIDAR points in the scene, lighting and colors of the scene, etc.).
MNIH does not specifically disclose, but CHEN discloses, (b) for each frame F1: (i) generating an image-feature tensor T1,img by applying a Residual-Network (ResNet) convolutional backbone (see ¶[0041] and [0228]; feature vector outputs from image feature extract attributes. See also ¶[0007] and [0245]; ResNet (K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learning for image recognition, arXiv preprint arXiv:1512.03385, 2015), Inception-ResNet-V2 (C. Szegedy, S. loffe, and V. Vanhoucke, Inception-v4, inception-resnet and the impact of residual connections on learning, CoRR, abs/1602.07261, 2016), to improve the capability and generality of visual feature extraction and hence enhance the accuracy of classification or regression) followed by two fully connected layers with rectified-linear-unit activations (see ¶[0234] and Fig. 3; rectified linear units);
(ii) in parallel, generating a meta-feature tensor T1,meta by applying the metadata vector to a three-layer fully connected meta model with rectified-linear-unit activations (see ¶[0041] and [0228]; feature vector outputs from image feature extract attributes. See also ¶[0007] and [0245]; ResNet (K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learning for image recognition, arXiv preprint arXiv:1512.03385, 2015), Inception-ResNet-V2 (C. Szegedy, S. loffe, and V. Vanhoucke, Inception-v4, inception-resnet and the impact of residual connections on learning, CoRR, abs/1602.07261, 2016), to improve the capability and generality of visual feature extraction and hence enhance the accuracy of classification or regression. See again ¶[0234]; a nonlinear activation on the output of fully connected layers);
(iii) concatenating T1,img and T1,meta into a tensor T1,cat and processing T1,cat through a merged model comprising three fully connected layers (see ¶[0080]; fine tune the deep model and retrain the fully connected layers against an objective function).
HARVEY does not specifically disclose, but NGUYEN discloses, a DropOut layer that randomly deactivates network units during training to reduce over-fitting (see ¶[0186]; dropout is applied at convolutional layers to reduce overfitting with randomly deactivating nodes from the network within a certain probability), and a Sigmoid output layer (see Fig. 3; FC + sigmoid).
The combination of HARVEY AND NGUYEN does not specifically disclose, but TAHA discloses, to yield class-probability logits L1 (see abstract; determine classification logits by applying a convolutional neural network).
(iv) deriving, from L1, machine-readable detection records that include brand- location coordinates, asset-placement type, scene label (see col 21, ln 34-49; the metadata may include labeling information 722, such as the labels for the object, probability distributions of the labels, classification score of the labels, etc. For another example, the metadata may include bounding box information 724 for that object, such as locations, distances (such as distance from the autonomous vehicle 110), dimensions, elevations, etc. For still another example, the metadata may include other object information 726, such as number of LIDAR points, colors in the object, lighting on the object, etc. For yet another example, the metadata may include scene information 728, such as the scene number, timestamp of the scene, total number of LIDAR points in the scene, lighting and colors of the scene, etc.).
HARVEY further discloses , and an active-versus-passive exposure flag (see ¶[0065]; a clickstream may be defined by digitized records produced by active people-metering technology and passive-metering technology).
HARVEY does not explicitly disclose, but NGUYEN discloses (c) completing steps (b)(i)-(b)(iv) for each frame within an interval less than a frame period of the video stream, achieving real-time inference (see ¶[0220]-[0222]; use the model in real time with real time
prediction capabilities);
(d) aggregating the per-frame detection records to compute sponsorship-exposure metrics that include at least brand exposure, asset exposure, scene-type exposure (see col 21, ln 34-49; the metadata may include labeling information 722, such as the labels for the object, probability distributions of the labels, classification score of the labels, etc. For another example, the metadata may include bounding box information 724 for that object, such as locations, distances (such as distance from the autonomous vehicle 110), dimensions, elevations, etc. For still another example, the metadata may include other object information 726, such as number of LIDAR points, colors in the object, lighting on the object, etc. For yet another example, the metadata may include scene information 728, such as the scene number, timestamp of the scene, total number of LIDAR points in the scene, lighting and colors of the scene, etc.)
HARVEY further discloses active exposure, and passive exposure (see ¶[0065]; a clickstream may be defined by digitized records produced by active people-metering technology and passive-metering technology).
HARVEY does not specifically disclose, but NORRIS discloses,, the active exposure and passive exposure differentiating between a logo that is digitally inserted into a frame from a logo that is captured in an original recording of the digital video stream (see ¶[0004]; signatures in the image can be used to detect fake images, which can be categorized into either passive or active approach. In the passive approach, imaging artifacts due to lens distortion, PRNU, and compression can be used to authenticate an image. An imperceptible watermark can be used in the passive approach).
HARVEY further discloses (e) supplying the sponsorship-exposure metrics to a client device (see ¶[0003] and [0007]; advertisers need to know which consumers view their commercials. Provide advertisers and other users with accurate measurements of the efficacy of their media advertising campaigns while promoting and protecting consumer privacy. See also ¶[0071] and [0077]; the advertiser can know which advertising has worked and which has not worked. Matched data can be made accessible in real-time or in batch to interested parties such as advertisers. See also ¶[0016] and Fig. 8; implement the service on a computing device).
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors. It would have been obvious to identify the logos using a neural network as taught by KATZ in the system executing the method of HARVEY with the motivation to apply a neural network to predict the presence of logos in images and measure the effectiveness of advertising.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]) based on data provided in images (see ¶[0195]). DOWDALL discloses perception visualization that includes metadata with labels for an object, bounding box information, dimensions, and scene information. It would have been obvious to include the metadata as taught by DOWDALL in the system executing the method of HARVEY with the motivation to label and provide information on objects in images (see DOWDALL abstract).
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). CHEN discloses predicting garment attributes using a neural network for analysis of online retail and shopping (see abstract and ¶[0001]-[0002]). It would have been obvious for one of ordinary skill in the art at the time of invention to use the neural network as taught by CHEN with the motivation to evaluate video media and prevent overfitting of the neural network.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). NGUYEN discloses deep learning with a neural network using dropout to reduce overfitting. It would have been obvious to use dropout as taught by NGUYEN in the system executing the method of HARVEY with the motivation to reduce overfitting in a neural network.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). NGUYEN discloses deep learning with a neural network using dropout to reduce overfitting. TAHA discloses visual recognition through neural network analysis of images to determine classification logits. It would have been obvious to include the classification logits as taught by TAHA in the system executing the method of HARVEY and NGUYEN with the motivation to measure the effect of advertising in image based media using a neural network.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). NORRIS discloses detecting fake images and videos using passive or active approaches. It would have been obvious for one of ordinary skill in the art at the time of invention to include fake image detection as taught by NORRIS in the system executing the method of HARVEY with the motivation to distinguish true positives and false positives in a detection network (see NORRIS ¶[0089] and HARVEY ¶[0069]).
Claim(s) 8, 9, and 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 20170140270 A1 to Mnih et al. (hereinafter ‘MNIH’) in view of US 20110288907 A1 to HARVEY et al., US 20180082123 A1 to KATZ et al., US 20220084223 A1 to NORRIS et al., US 20200320769 A1 to CHEN et al., US 20200302223 A1 to DUTTA et al., US 20200234088 A1 to TAHA et al., and US 10699167 B1 to DOWDALL et al.
Claim 8 (Currently Amended)
MNIH discloses an intelligent secure networked messaging system configured by at least one processor to execute instructions stored in memory (see ¶[0007]-[0008] and [0087]-[0088]; a programmable processor. Computer programs configured to perform operations or action means. The computer storage medium can be a memory device), the system comprising: a data retention system (see ¶[0088]; a database management system).
MNIH does not specifically disclose, but HARVEY discloses, configured to store temporal sponsorship data (see ¶[0062]-[0063] and [0065]; the sum of purchaser ratings in a time period. Time-stamped media exposure events) with encrypted sponsor identifiers (see ¶[0053], [0073]; a watermark embedding a commercial code or program code).
MNIH further discloses and an analytics system (see ¶[0007]; a data processing apparatus), the analytics system performing asynchronous processing with a computing device (see abstract; asynchronous deep reinforcement learning) and the analytics system communicatively coupled to a deep neural network (see abstract; asynchronous deep reinforcement learning in a deep neural network).
MNIH does not specifically disclose, but HARVEY discloses, the deep neural network configured to: perform real-time analysis of a video source media (see ¶[0200]-[0201]; video advertising networks) including a sponsor message (see abstract and ¶[0071]; an advertising measurement system. A touchpoint includes event sponsorships) and a source entity ID (see ¶[0053], [0073]; a watermark embedding a commercial code or program code).
MNIH does not specifically disclose, but KATZ discloses, each frame being processed by the processor within an inter-frame interval of the video source media (see ¶[0095] and [0101]-[0103] and Fig. 8; a user defines a bounding box. The computing system may track movement of the object within the box and store estimated bounding box information with respect to subsequent frames. Overlays may include visual bounding boxes around sponsor logos recognized by the above techniques, which may follow movement of the logo from frame to frame, as well as graphical indicators of the media value for that exposure),
the real-time analysis comprising: generating one or more brand locations and brand sizes within the video source media with [sic] least one brand detection neural networks (see ¶[0024]; a neural network classifier for image object recognition), the brand detection neural networks trained to detect a sponsor message (see ¶[0027] and [0054]; the size of the logo within the frame and where the logo appears. Media percentage is measured based on clarity, visibility, size, and placement. See also ¶[0146]; “Feeling Anxious? Grab and CandyBar”).
MNIH does not specifically disclose, but HARVEY discloses, using dual-modality feature extraction (see ¶[0041] and [0228]; feature vector outputs from image feature extract attributes).
MNIH does not specifically disclose, but KATZ discloses, executing a sports-specific asset detection model, the sports-specific asset detection model including a plurality of neural networks each trained for a specific sport (see abstract and ¶[0024]; a first set of classification models that are each trained to identify at least one type of scene associated with one or more sports, then features of the media content may be provided to a second set of classification models trained to identify at least one object associated with one or more sports. A neural network as a classifier).
MNIH does not specifically disclose, but NORRIS discloses the plurality of neural networks comprising a detection neural network used by an active/passive detection module, trained to differentiate a logo that is digitally inserted into a frame from a logo that is captured in an original recording of the video source media see ¶[0004]; signatures in the image can be used to detect fake images, which can be categorized into either passive or active approach. In the passive approach, imaging artifacts due to lens distortion, PRNU, and compression can be used to authenticate an image. An imperceptible watermark can be used in the passive approach).
MNIH does not specifically disclose, but CHEN discloses the detection neural network comprising: a base image model implementing a ResNet convolutional backbone generating a base image output tensor (see ¶[0041] and [0228]; feature vector outputs from image feature extract attributes. See also ¶[0007] and [0245]; ResNet (K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learning for image recognition, arXiv preprint arXiv:1512.03385, 2015), Inception-ResNet-V2 (C. Szegedy, S. loffe, and V. Vanhoucke, Inception-v4, inception-resnet and the impact of residual connections on learning, CoRR, abs/1602.07261, 2016), to improve the capability and generality of visual feature extraction and hence enhance the accuracy of classification or regression) from pixel data (see abstract and ¶[0028]; an image. High-resolution photos);
a base meta model implementing a multi-layer fully connected network generating a base meta model output feature tensor (see ¶[0045]; fully connected layers for attribute classification ).
MNIH does not specifically disclose, but DUTTA discloses, a merged model comprising three fully connected layers paired with a Rectified Linear Unit, a dropout layer, and a Sigmoid activation to an output (see ¶[0298]; activation functions such as ReLU, sigmoid and dropout can be used in a neural network).
MNIH does not specifically disclose, but HARVEY discloses, the merged model receiving a concatenated single feature tensor, the concatenated single feature tensor being the base image output tensor and the base meta model output feature tensor merged together (see ¶[0041]-[0045]; visual features are vectorized and concatenated, and passed to fully-connected layers for attribute classification), the merged model reducing overfitting in the detection neural network (see ¶[0245] and [0410]; we did not want to overfit the data, and thus wanted to have a lower number of parameters. Prevent neural network from overfitting. See also ¶[0265]; an additional fully-connected layer for dimensionality reduction through by vector concatenation).
MNIH does not specifically disclose, but CHEN discloses, the parameter-controlled regularization dynamically adjusts network weights based on training loss convergence (see ¶[0037]; minimize overall loss, defined as a weighted sum of the loss function, on each attribute).
MNIH does not specifically disclose, but HARVEY discloses, receive a video source media at a first input that includes a sponsor message (see ¶[0200]-[0201]; video advertising networks) and a source entity ID (see ¶[0053], [0073]; a watermark embedding a commercial code or program code).
MNIH further discloses, at an input layer (see ¶[0005] and [0049] & claims 1 and 6; deep neural networks include one or more hidden layers in addition to an output layer. Process neural network inputs using the deep neural network. Predict an output for a received input).
MNIH does not specifically disclose, but HARVEY discloses, process the first input by one or more hidden layers trained to detect a sponsor message generating a first output (see ¶[0024], [0027]. And [0054]; a neural network classifier for image object recognition. In the context of object recognition in digital images, classifiers may be used to determine a likelihood that a particular image object (e.g., a visual representation of an object, or a company logo) is included or depicted in an image).
The combination of MNIH and HARVEY does not explicitly disclose, but TAHA discloses, comprising class-probability logits (see abstract; determine classification logits by applying a convolutional neural network).
MNIH further discloses transmit the first output to an output layer (see again ¶[0005]; include one or more hidden layers in addition to an output layer).
MNIH does not specifically disclose, but KATZ discloses, configured to generate machine-readable detection records (see claim 1; neural network trained to identify a logo associated with sponsors. Determine that a first sponsor logo appears in the media content based at least in part on output of the plurality of neural networks associated with the one or more sponsors); map the first output to a sponsor (see again claim 1; neural network trained to identify a logo associated with sponsors. Determine that a first sponsor logo appears in the media content based at least in part on output of the plurality of neural networks associated with the one or more sponsors).
HARVEY does not specifically disclose, but DOWDALL discloses, output an analyzed video source media with a bounding box around the detected asset indicating the location and size of the asset (see col 21, ln 34-49; the metadata may include labeling information 722, such as the labels for the object, probability distributions of the labels, classification score of the labels, etc. For another example, the metadata may include bounding box information 724 for that object, such as locations, distances (such as distance from the autonomous vehicle 110), dimensions, elevations, etc. For still another example, the metadata may include other object information 726, such as number of LIDAR points, colors in the object, lighting on the object, etc. For yet another example, the metadata may include scene information 728, such as the scene number, timestamp of the scene, total number of LIDAR points in the scene, lighting and colors of the scene, etc.).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. HARVEY discloses using a neural network to analyze video for television targeting. It would have been obvious to include the prediction method as taught by MNIH in the system executing the method of HARVEY with the motivation to make predictions for marketing purposes (see HARVEY ¶[0005]).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors. It would have been obvious to identify the logos using a neural network as taught by KATZ in the system executing the method of HARVEY and MNIH with the motivation to apply a neural network to predict the presence of logos in images and measure the effectiveness of advertising to make marketing investments.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). NORRIS discloses detecting fake images and videos using passive or active approaches. It would have been obvious for one of ordinary skill in the art at the time of invention to include fake image detection as taught by NORRIS in the system executing the method of HARVEY with the motivation to distinguish true positives and false positives in a detection network (see NORRIS ¶[0089] and HARVEY ¶[0069]).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. CHEN discloses predicting garment attributes using a neural network for analysis of online retail and shopping (see abstract and ¶[0001]-[0002]). It would have been obvious for one of ordinary skill in the art at the time of invention to use the neural network as taught by CHEN in the system executing the method of MNIH with the motivation to predict garment attributes.
HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). DUTTA discloses artificial intelligence that uses neural networks with layers comprised of activation functions including ReLU, sigmoid, and dropout. It would have been obvious for one of ordinary skill in the art at the time of invention to include the activation functions as taught by DUTTA in the system executing the method of HARVEY with the motivation to use a neural network to measure the effect of advertising on viewership.
MNIH discloses asynchronous deep reinforcement learning in a neural network with input layers, hidden layers, and output layers to make a prediction based on received input (see abstract). HARVEY discloses using consumer purchase behavior for television targeting that uses a neural network to measure the effect of advertising on viewership by measuring exposure to media and marketing communications (see ¶[0005]). TAHA discloses visual recognition through neural network analysis of images to determine classification logits. It would have been obvious to include the classification logits as taught by TAHA in the system executing the method of MNIH and HARVEY with the motivation to measure the effect of advertising in image based media using a neural network.
MNIH discloses asynchronous deep reinforcement learning in a neural network with input layers, hidden layers, and output layers to make a prediction based on received input (see abstract). DOWDALL discloses perception visualization that includes metadata with labels for an object, bounding box information, dimensions, and scene information; with object detection models including a neural net model (see col 9, ln 16-31). It would have been obvious to include the metadata as taught by DOWDALL in the system executing the method of MNIH with the motivation to label and provide information on objects in images using a neural network (see DOWDALL abstract).
Claim 9 (Original)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL discloses the intelligent secure networked messaging system as set forth in claim 8.
MNIH does not explicitly disclose, but KATZ discloses, further comprising the sponsor is an outcome (see claim 1; neural network trained to identify a logo associated with sponsors. Determine that a first sponsor logo appears in the media content based at least in part on output of the plurality of neural networks associated with the one or more sponsors).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors. It would have been obvious to identify the logos using a neural network as taught by KATZ in the system executing the method of MNIH with the motivation to apply a neural network to predict the presence of logos in images.
Claim 14 (Previously Presented)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL discloses the intelligent secure networked messaging system as set forth in claim 8.
MNIH does not explicitly disclose, but HARVEY discloses, wherein the video source media includes a video from social media (see ¶[0065] and [0071]; a touchpoint is all means or media by which consumers may be influenced by marketing or advertising, including social media).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. HARVEY discloses using a neural network to analyze video for television targeting. It would have been obvious to include the prediction method as taught by MNIH in the system executing the method of HARVEY with the motivation to make predictions for marketing purposes (see HARVEY ¶[0005]).
Claim(s) 10 and 11 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 20170140270 A1 to MNIH et al. in view of US 20110288907 A1 to HARVEY et al., US 20180082123 A1 to KATZ et al., US 20220084223 A1 to NORRIS et al., US 20200320769 A1 to CHEN et al., US 20200302223 A1 to DUTTA et al., US 20200234088 A1 to TAHA et al., and US 10699167 B1 to DOWDALL et al. as applied to claims 8 and 9 above, and further in view of US 20210158036 A1 to Huber (hereinafter ‘HUBER’).
Claim 10 (Original)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL discloses the intelligent secure networked messaging system as set forth in claim 9.
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL does not specifically disclose, but HUBER discloses, further comprising the first outcome being transmitted to the input layer as input (see ¶[0063]; a training process for generating one or more deep neural network layers that uses a feedback loop for adjusting weights until the output layer drives output data that accurately classifies input data. The output data presenting a class for the input data can be represented by a similarity score).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. HUBER discloses deep neural network layers using a feedback loop to adjust weights using an output that is compared to an input to train the learning algorithm. It would have been obvious to train an output by iteratively feeding it into input with the motivation to reduce error and improve the accuracy of the algorithm.
Claim 11 (Original)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, DOWDALL, and HUBER discloses the intelligent secure networked messaging system of claim 10.
MNIH does not specifically disclose, but HUBER discloses, further comprising the second outcome being transmitted to the input layer (see ¶[0063]; a training process for generating one or more deep neural network layers that uses a feedback loop for adjusting weights until the output layer drives output data that accurately classifies input data. The output data presenting a class for the input data can be represented by a similarity score).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. HUBER discloses deep neural network layers using a feedback loop to adjust weights using an output that is compared to an input to train the learning algorithm. It would have been obvious to train an output by iteratively feeding it into input with the motivation to reduce error and improve the accuracy of the algorithm.
Claim(s) 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 20170140270 A1 to MNIH et al. in view of US 20110288907 A1 to HARVEY et al., US 20180082123 A1 to KATZ et al., US 20220084223 A1 to NORRIS et al., US 20200320769 A1 to CHEN et al., US 20200302223 A1 to DUTTA et al., US 20200234088 A1 to TAHA et al., and US 10699167 B1 to DOWDALL et al. as applied to claim 8 above, and further in view of US 20210248629 A1 to Sullivan et al. (hereinafter ‘SULLIVAN’).
Claim 15 (Previously Presented)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL discloses the intelligent secure networked messaging system of claim 8.
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL does not specifically disclose, but SULLIVAN discloses, wherein the source media includes a text (see ¶[0037]; a media impression is defined as an occurrence of access and/or exposure to media. Monitor for media impressions of video and text).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors and count the number of media impressions (see ¶[0120]). SULLIVAN discloses media impressions defined as occurrence of access and/or exposure to media types that includes video and text. It would have been obvious to include the video and text as taught by SULLIVAN in the system executing the method of MNIH and KATZ with the motivation to identify and predict media impressions of different media types.
Claim(s) 16-18 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 20170140270 A1 to MNIH et al. in view of US 20110288907 A1 to HARVEY et al., US 20180082123 A1 to KATZ et al., US 20220084223 A1 to NORRIS et al., US 20200320769 A1 to CHEN et al., US 20200302223 A1 to DUTTA et al., US 20200234088 A1 to TAHA et al., and US 10699167 B1 to DOWDALL et al. as applied to claim 8 above, and further in view of US 20150317670 A1 to Cavander et al. (hereinafter ‘CAVANDER’).
Claim 16 (Previously Presented)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL discloses the intelligent secure networked messaging system of claim 8.
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, and DOWDALL does not specifically disclose but CAVANDER discloses, further comprising the source media including a sponsor message (see ¶[0016] and [0023]; a marketing element includes newspaper and direct mail advertisements. Marketing elements can include a time and duration of exposure and a message campaign).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors and count the number of media impressions (see ¶[0120]). CAVANDER discloses measuring exposure to messages and content from advertisers. It would have been obvious to measure exposure to messages and content from advertisers as taught by CAVANDER in the system executing the method of MNIH and KATZ with the motivation to count media impressions.
Claim 17 (Original)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, DOWDALL, and CAVANDER discloses the intelligent secure networked messaging system as set forth in claim 16.
MNIH does not specifically disclose, but CAVANDER discloses, wherein the sponsor message includes a brand name (see ¶[0016]; marketing elements can include a brand).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors and count the number of media impressions (see ¶[0120]). CAVANDER discloses measuring exposure to messages and content from advertisers regarding a brand. It would have been obvious to measure exposure to messages and content from advertisers as taught by CAVANDER in the system executing the method of MNIH and KATZ with the motivation to count media impressions.
Claim 18 (Original)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, DOWDALL, and CAVANDER discloses the intelligent secure networked messaging system as set forth in claim 16.
MNIH does not specifically disclose, but CAVANDER discloses, wherein the sponsor message includes a logo (see ¶[0016]; trademarks or logos).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors and count the number of media impressions (see ¶[0120]). CAVANDER discloses measuring exposure to messages and content from advertisers regarding brands, including trademark and logo impressions. It would have been obvious to measure exposure to messages and content from advertisers as taught by CAVANDER in the system executing the method of MNIH and KATZ with the motivation to count media impressions.
Claim(s) 19 is/are rejected under 35 U.S.C. 103 as being unpatentable over US 20170140270 A1 to MNIH et al. in view of US 20110288907 A1 to HARVEY et al., US 20180082123 A1 to KATZ et al., US 20220084223 A1 to NORRIS et al., US 20200320769 A1 to CHEN et al., US 20200302223 A1 to DUTTA et al., US 20200234088 A1 to TAHA et al., US 10699167 B1 to DOWDALL et al., and US 20150317670 A1 to Cavander et al.as applied to claims 8 and 16 above, and further in view of US 20160203225 A1 to Alonso et al. (hereinafter ‘ALONSO’).
Claim 19 (Original)
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, DOWDALL, and CAVANDER discloses the intelligent secure networked messaging system as set forth in claim 16.
The combination of MNIH, HARVEY, KATZ, NORRIS, CHEN, DUTTA, TAHA, DOWDALL, and CAVANDER does not specifically disclose, but ALONSO discloses, wherein the sponsor message includes a slogan (see ¶[0034]; evaluate a post to determine whether it includes a slogan).
MNIH discloses asynchronous deep reinforcement learning with input layers, hidden layers, and output layers to make a prediction based on received input. KATZ discloses machine learning models including a neural network trained to identify a logo associated with sponsors and count the number of media impressions (see ¶[0120]). ALONSO discloses evaluating an online post to determine whether it includes a slogan to determine a likelihood it is associated with an event. It would have been obvious to include the slogan recognition as taught by ALONSO in the system executing the method of MNIGH and KATZ with the motivation to determine media impressions associated with a logo or slogan.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RICHARD N SCHEUNEMANN whose telephone number is (571)270-7947. The examiner can normally be reached M-F 9am-5pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Patricia Munson can be reached at 571-270-5396. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/RICHARD N SCHEUNEMANN/Primary Examiner, Art Unit 3624
Prosecution Insights