DETAILED ACTION
Claims 1-20 are pending in this action.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Interpretation
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
As explained in MPEP § 2181, subsection I, claim limitations that meet the following three-prong test will be interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph:
(A) the claim limitation uses the term “means” or “step” or a term used as a substitute for “means” that is a generic placeholder (also called a nonce term or a non-structural term having no specific structural meaning) for performing the claimed function;
(B) the term “means” or “step” or the generic placeholder is modified by functional language, typically, but not always linked by the transition word “for” (e.g., “means for”) or another linking word or phrase, such as “configured to” or “so that”; and
(C) the term “means” or “step” or the generic placeholder is not modified by sufficient structure, material, or acts for performing the claimed function.
Use of the word “means” (or “step”) in a claim with functional language creates a rebuttable presumption that the claim limitation is to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites sufficient structure, material, or acts to entirely perform the recited function.
Absence of the word “means” (or “step”) in a claim creates a rebuttable presumption that the claim limitation is not to be treated in accordance with 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph. The presumption that the claim limitation is not interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is rebutted when the claim limitation recites function without reciting sufficient structure, material or acts to entirely perform the recited function.
Claim limitations in this application that use the word “means” (or “step”) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action. Conversely, claim limitations in this application that do not use the word “means” (or “step”) are not being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, except as otherwise indicated in an Office action.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitations are:
Processing unit of claim 9.
Electronic device of claims 9 and 17.
Because this/these claim limitation(s) is/are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, it/they is/are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have this/these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 2, 10 and 18 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. The claims recite that the “first content data comprises at least one of: image data associated with the target media content; text data associated with the target media content”, which renders that claim indefinite because it is unclear whether the first content data requires both image data and text data. For examination purposes, the examiner is interpreting this limitation as the first data comprising either text data or image data associated with the media. Applicant is encouraged to amend to clarify this limitation.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Iscen (US 20240320493 A1) and Zhuo (US 20180012236 A1).
Regarding claim 1 Iscen discloses; A method for classifying a media content, comprising:
determining a set of target features of a target media content based on first content data of the target media content (Iscen, [0029] an input of images or videos is received by one or more classification models to generate a set of features based on the input data);
and processing, using a first classification model, the set of target features of the media content to generate classification information for the target media content (Iscen, [0029] the one more classification model (first model) processes the input data to generate features, the features are then processed by the model to generate a classification output),
PNG
media_image1.png
468
408
media_image1.png
Greyscale
(Iscen, [0029], emphasis added)
the first classification model being trained through distilling a second classification model (Iscen, [0033] a teacher model (first model) may be trained and distilled using a student model (second model)),
PNG
media_image2.png
252
406
media_image2.png
Greyscale
(Iscen, [0033])
the second classification model being configured to generate classification information of a first training media content based on both a first set of features and a second set of features of the first training media content (Iscen, [0102]-[0104] and figure 3, teaches that the training process takes place in two stages where first one or more teacher models are trained on one input (first training media content), these one or more teacher models generate one or more feature sets (first and second features) from the input data, the outputs from this step are then distilled and passed to one or more student models (second classification model) to generate a classification output from the inputs, figure 3 shows the outputs of two teacher models being distilled and input into a classifier to obtain a classifier output which would then be based on these two sets of features from the media input),
PNG
media_image3.png
366
854
media_image3.png
Greyscale
(Iscen, Figure 3)
the first set of features being determined based on second content data of the first training media content (Iscen, [0102]-[0104] the training takes place in two stages where the first stage input may be different from the second, therefore there is at least a first and a second input, and at least a first and second set of features being generated from each input, figure 3 displays this, where inputs 312 and 332 have multiple sets of features generated from them),
[and the second set of features being determined based on interaction data associated with the first training media content.]
Iscen fails to teach; and the second set of features being determined based on interaction data associated with the first training media content.
However, in the same field of endeavor of media content classification, Zhuo teaches; and the second set of features being determined based on interaction data associated with the first training media content (Zhuo, [0005] A classifier is provided content data from social media (first training media content) and [0006] one or more features are extracted from the data (at least a first and second feature set) where one of the features may be [0007] interaction content such as a like or share).
The combination of Iscen and Zhuo would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. Iscen teaches a method of input content distillation using multiple models to extract features and classify them. The motivation to add the interaction-based feature extraction of Zhuo to this method of Iscen is that analyzing and classifying interaction data of a user or multiple users helps to provide content to them that they are more likely to engage with. (Zhuo, [0002]-[0009])
Regarding claim 2 the combination of Iscen and Zhuo teaches; The method of claim 1, wherein first content data comprises at least one of: image data associated with the target media content (Iscen, [0073] the input may be an image to be classified or in some cases [0072] an image which an embedding must be generated for, the examiner is interpreting this claim as only requiring one of the two provided data types to satisfy the limitation because it states at least one of the type is required, not both);
text data associated with the target media content (Zhuo, [0009] the data may be images or text associated with the images or posts).
The combination of Iscen and Zhou would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. The motivation for the addition of the text data associated with media content of Zhuo is that this allows the model to be trained on both an image from a post as well any caption data to help provide more tailored social media content. (Zhuo, [0001]-[0011])
Regarding claim 3 the combination of Iscen and Zhuo teaches; The method of claim 1, wherein the first classification model is trained through: determining a first loss based on a first difference between first classification information and reference classification information of a second training media content (Iscen, [0029] the model uses a cross-entropy loss function which determines loss by assessing the difference between a prediction and the training data/ground truth, [0059] loss is computed for each model, therefore loss would be determined for each of the first and second models, based on the training data),
the first classification information being generated by the first classification model (Iscen, Figure 3, the input data is input into the teacher models (first model) and then a classification output is generated);
PNG
media_image4.png
366
854
media_image4.png
Greyscale
(Iscen, Figure 3, emphasis added)
determining a second loss based on a second difference between the first classification information and second classification information for the second training media content (Iscen, [0094] another loss term may be computed looking at the differences between the classification from the teacher model (first model) and the student model (second model), where the two output classifications (first and second classifications) are compared),
PNG
media_image5.png
300
322
media_image5.png
Greyscale
(Iscen, [0093]-[0094])
the second classification information being generated by the second classification model (Iscen, [0051] a trained student model (second model) outputs a classification (second classification));
determining a target loss based on the first loss and the second loss (Iscen, [0096] a distillation loss term may be generated (target loss) to minimize loss and encourage the student and teacher outputs to be the same, the equation shown uses both the first and second loss values to generate this distillation loss (target loss));
and training the first classification model based on the target loss (Iscen, [0096] the training process can take the distillation loss into account to train the teacher model (first classification model)).
PNG
media_image6.png
48
322
media_image6.png
Greyscale
PNG
media_image7.png
286
334
media_image7.png
Greyscale
(Iscen, [0096])
Regarding claim 4 the combination of Iscen and Zhuo teaches; The method of claim 3, wherein determining a target loss based on the first loss and the second loss comprises: determining weight information based on a confidence for the second loss classification information (Iscen, [0087] class confidence scores are output as zi, [0088]-[0089] this function zi is a parameter in one of the loss functions which would weight/scale the parameters accordingly);
and determining a weighted sum of the first loss and the second loss according to the weight information (Iscen, [0094] the two loss terms are balanced where weight T can be adjusted to balance the weighted sum shown in the equation).
PNG
media_image8.png
226
316
media_image8.png
Greyscale
(Iscen, [0094])
Regarding claim 5 the combination of Iscen and Zhuo teaches; The method of claim 4, wherein a target weight corresponding to the second loss is proportional to the confidence (Iscen, [0087]-[0089] the confidence score output of the model is factored into the loss equation, which scales/weights the equation according the confidence, therefore the target weight would be related to or proportional to the confidence).
Regarding claim 6 the combination of Iscen and Zhuo teaches; The method of claim 5, wherein the target weight is determined according to a preset function of the confidence (Iscen, [0087] the model outputs confidence scores according the function below, this function is denoted as variable zi, Where this variable is included the equation for the loss and therefore is analogous to a target weight, this is also analogous to what applicant has defined the target weight as in paragraphs [0051]-[0052] of the specification).
Regarding claim 7 the combination of Iscen and Zhuo teaches; The method of claim 4, further comprising: determining loss information for the second training media content of the second classification model (Iscen, [0094] loss is computed for both the teacher models (first loss for first classification models) and for the student models (second loss for second classification models));
and determining the confidence based on the loss information (Iscen, [0094] in the equation shown in this paragraph, the output confidence (zi) is scaled and adjusted based on the loss information l).
Regarding claim 8 the combination of Iscen and Zhuo teaches; The method of claim 1, wherein a feature related to interaction data for the target media content is omitted from being input to the first classification model (Zhuo, [0030]-[0031] and [0038] the system uses a minimum threshold to determine if content data meets the criteria for being “click-bait” and all data used in training the model must meet this threshold, therefore the media that does not meet this interaction based metric would be omitted from being input into the model).
The combination of Iscen and Zhuo would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. Iscen teaches a method of input content distillation using multiple models to extract features and classify them. The motivation to add the interaction-based filtering of the content of Zhuo would improve the system by filtering less relevant content from the training data to improve the relevance of the training data set. This would allow for an increase in accuracy and relevance of the results output by the trained model. (Zhuo, [0030]-[0031], [0038])
Regarding claim 9 the combination of Iscen and Zhuo teaches; An electronic device, comprising:
at least one processing unit (Iscen, [0008] the system has one or more processors);
and at least one memory coupled to the at least one processing unit and storing instructions executable by the at least one processing unit, the instructions, upon execution by the at least one processing unit, causing the electronic device to perform actions comprising (Iscen, [0008] the system has one or more processors coupled to non-transitory medium to execute and store instructions):
determining a set of target features of a target media content based on first content data of the target media content (Iscen, [0029] an input of images or videos is received by one or more classification models to generate a set of features based on the input data);
and processing, using a first classification model, the set of target features of the media content to generate classification information for the target media content (Iscen, [0029] the one more classification model (first model) processes the input data to generate features, the features are then processed by the model to generate a classification output),
the first classification model being trained through distilling a second classification model (Iscen, [0033] a teacher model (first model) may be trained and distilled using a student model (second model)),
the second classification model being configured to generate classification information of a first training media content based on both a first set of features and a second set of features of the first training media content (Iscen, [0102]-[0104] and figure 3, teaches that the training process takes place in two stages where first one or more teacher models are trained on one input (first training media content), these one or more teacher models generate one or more feature sets (first and second features) from the input data, the outputs from this step are then distilled and passed to one or more student models (second classification model) to generate a classification output from the inputs, figure 3 shows the outputs of two teacher models being distilled and input into a classifier to obtain a classifier output which would then be based on these two sets of features from the media input),
the first set of features being determined based on second content data of the first training media content (Iscen, [0102]-[0104] the training takes place in two stages where the first stage input may be different from the second, therefore there is at least a first and a second input, and at least a first and second set of features being generated from each input, figure 3 displays this, where inputs 312 and 332 have multiple sets of features generated from them),
and the second set of features being determined based on interaction data associated with the first training media content (Zhuo, [0005] A classifier is provided content data from social media (first training media content) and [0006] one or more features are extracted from the data (at least a first and second feature set) where one of the features may be [0007] interaction content such as a like or share).
The combination of Iscen and Zhuo would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. Iscen teaches a method of input content distillation using multiple models to extract features and classify them. The motivation to add the interaction-based feature extraction of Zhuo to this method of Iscen is that analyzing and classifying interaction data of a user or multiple users helps to provide content to them that they are more likely to engage with. (Zhuo, [0002]-[0009])
Regarding claim 10 the combination of Iscen and Zhuo teaches; The electronic device of claim 9, wherein first content data comprises at least one of: image data associated with the target media content (Iscen, [0073] the input may be an image to be classified or in some cases [0072] an image which an embedding must be generated for, the examiner is interpreting this claim as only requiring one of the two provided data types to satisfy the limitation because it states at least one of the type is required, not both);
text data associated with the target media content (Zhuo, [0009] the data may be images or text associated with the images or posts).
The combination of Iscen and Zhou would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. The motivation for the addition of the text data associated with media content of Zhuo is that this allows the model to be trained on both an image from a post as well any caption data to help provide more tailored social media content. (Zhuo, [0001]-[0011])
Regarding claim 11 the combination of Iscen and Zhuo teaches; The electronic device of claim 9, wherein the first classification model is trained through: determining a first loss based on a first difference between first classification information and reference classification information of a second training media content (Iscen, [0029] the model uses a cross-entropy loss function which determines loss by assessing the difference between a prediction and the training data/ground truth, [0059] loss is computed for each model, therefore loss would be determined for each of the first and second models, based on the training data),
the first classification information being generated by the first classification model (Iscen, Figure 3, the input data is input into the teacher models (first model) and then a classification output is generated);
determining a second loss based on a second difference between the first classification information and second classification information for the second training media content (Iscen, [0094] another loss term may be computed looking at the differences between the classification from the teacher model (first model) and the student model (second model), where the two output classifications (first and second classifications) are compared),
the second classification information being generated by the second classification model (Iscen, [0051] a trained student model (second model) outputs a classification (second classification));
determining a target loss based on the first loss and the second loss (Iscen, [0096] a distillation loss term may be generated (target loss) to minimize loss and encourage the student and teacher outputs to be the same, the equation shown uses both the first and second loss values to generate this distillation loss (target loss));
and training the first classification model based on the target loss (Iscen, [0096] the training process can take the distillation loss into account to train the teacher model (first classification model)).
Regarding claim 12 the combination of Iscen and Zhuo teaches; The electronic device of claim 11, wherein determining a target loss based on the first loss and the second loss comprises: determining weight information based on a confidence for the second loss classification information (Iscen, [0087] class confidence scores are output as zi, [0088]-[0089] this function zi is a parameter in one of the loss functions which would weight/scale the parameters accordingly);
and determining a weighted sum of the first loss and the second loss according to the weight information (Iscen, [0094] the two loss terms are balanced where weight T can be adjusted to balance the weighted sum shown in the equation).
PNG
media_image8.png
226
316
media_image8.png
Greyscale
(Iscen, [0094])
Regarding claim 13 the combination of Iscen and Zhuo teaches; The electronic device of claim 12, wherein a target weight corresponding to the second loss is proportional to the confidence (Iscen, [0087]-[0089] the confidence score output of the model is factored into the loss equation, which scales/weights the equation according the confidence, therefore the target weight would be related to or proportional to the confidence).
Regarding claim 14 the combination of Iscen and Zhuo teaches; The electronic device of claim 13, wherein the target weight is determined according to a preset function of the confidence (Iscen, [0087] the model outputs confidence scores according the function below, this function is denoted as variable zi, Where this variable is included the equation for the loss and therefore is analogous to a target weight, this is also analogous to what applicant has defined the target weight as in paragraphs [0051]-[0052] of the specification).
Regarding claim 15 the combination of Iscen and Zhuo teaches; The electronic device of claim 12, wherein the actions further comprise: determining loss information for the second training media content of the second classification model (Iscen, [0094] loss is computed for both the teacher models (first loss for first classification models) and for the student models (second loss for second classification models));
and determining the confidence based on the loss information (Iscen, [0094] in the equation shown in this paragraph, the output confidence (zi) is scaled and adjusted based on the loss information l).
Regarding claim 16 the combination of Iscen and Zhuo teaches; The electronic device of claim 9, wherein a feature related to interaction data for the target media content is omitted from being input to the first classification model (Zhuo, [0030]-[0031] and [0038] the system uses a minimum threshold to determine if content data meets the criteria for being “click-bait” and all data used in training the model must meet this threshold, therefore the media that does not meet this interaction based metric would be omitted from being input into the model).
The combination of Iscen and Zhuo would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. Iscen teaches a method of input content distillation using multiple models to extract features and classify them. The motivation to add the interaction-based filtering of the content of Zhuo would improve the system by filtering less relevant content from the training data to improve the relevance of the training data set. This would allow for an increase in accuracy and relevance of the results output by the trained model. (Zhuo, [0030]-[0031], [0038])
Regarding claim 17 the combination of Iscen and Zhuo teaches; A non-transitory computer-readable storage medium, having a computer program stored thereon which, upon execution by an electronic device, causes the device to perform actions comprising (Iscen, [0008] the system has one or more processors coupled to non-transitory medium to execute and store instructions):
determining a set of target features of a target media content based on first content data of the target media content (Iscen, [0029] an input of images or videos is received by one or more classification models to generate a set of features based on the input data);
and processing, using a first classification model, the set of target features of the media content to generate classification information for the target media content (Iscen, [0029] the one more classification model (first model) processes the input data to generate features, the features are then processed by the model to generate a classification output),
the first classification model being trained through distilling a second classification model (Iscen, [0033] a teacher model (first model) may be trained and distilled using a student model (second model)),
the second classification model being configured to generate classification information of a first training media content based on both a first set of features and a second set of features of the first training media content (Iscen, [0102]-[0104] and figure 3, teaches that the training process takes place in two stages where first one or more teacher models are trained on one input (first training media content), these one or more teacher models generate one or more feature sets (first and second features) from the input data, the outputs from this step are then distilled and passed to one or more student models (second classification model) to generate a classification output from the inputs, figure 3 shows the outputs of two teacher models being distilled and input into a classifier to obtain a classifier output which would then be based on these two sets of features from the media input),
the first set of features being determined based on second content data of the first training media content (Iscen, [0102]-[0104] the training takes place in two stages where the first stage input may be different from the second, therefore there is at least a first and a second input, and at least a first and second set of features being generated from each input, figure 3 displays this, where inputs 312 and 332 have multiple sets of features generated from them),
and the second set of features being determined based on interaction data associated with the first training media content (Zhuo, [0005] A classifier is provided content data from social media (first training media content) and [0006] one or more features are extracted from the data (at least a first and second feature set) where one of the features may be [0007] interaction content such as a like or share).
The combination of Iscen and Zhuo would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. Iscen teaches a method of input content distillation using multiple models to extract features and classify them. The motivation to add the interaction-based feature extraction of Zhuo to this method of Iscen is that analyzing and classifying interaction data of a user or multiple users helps to provide content to them that they are more likely to engage with. (Zhuo, [0002]-[0009])
Regarding claim 18 the combination of Iscen and Zhuo teaches; The non-transitory computer-readable storage medium of claim 17, wherein first content data comprises at least one of: image data associated with the target media content; (Iscen, [0073] the input may be an image to be classified or in some cases [0072] an image which an embedding must be generated for, the examiner is interpreting this claim as only requiring one of the two provided data types to satisfy the limitation because it states at least one of the type is required, not both);
text data associated with the target media content (Zhuo, [0009] the data may be images or text associated with the images or posts).
The combination of Iscen and Zhou would have been obvious to one of ordinary skill in the art prior to the effective filing date of the presently claimed invention. The motivation for the addition of the text data associated with media content of Zhuo is that this allows the model to be trained on both an image from a post as well any caption data to help provide more tailored social media content. (Zhuo, [0001]-[0011])
Regarding claim 19 the combination of Iscen and Zhuo teaches; The non-transitory computer-readable storage medium of claim 17, wherein the first classification model is trained through: determining a first loss based on a first difference between first classification information and reference classification information of a second training media content (Iscen, [0029] the model uses a cross-entropy loss function which determines loss by assessing the difference between a prediction and the training data/ground truth, [0059] loss is computed for each model, therefore loss would be determined for each of the first and second models, based on the training data),
the first classification information being generated by the first classification model (Iscen, Figure 3, the input data is input into the teacher models (first model) and then a classification output is generated);
determining a second loss based on a second difference between the first classification information and second classification information for the second training media content (Iscen, [0094] another loss term may be computed looking at the differences between the classification from the teacher model (first model) and the student model (second model), where the two output classifications (first and second classifications) are compared),
the second classification information being generated by the second classification model (Iscen, [0051] a trained student model (second model) outputs a classification (second classification));
determining a target loss based on the first loss and the second loss (Iscen, [0096] a distillation loss term may be generated (target loss) to minimize loss and encourage the student and teacher outputs to be the same, the equation shown uses both the first and second loss values to generate this distillation loss (target loss));
and training the first classification model based on the target loss (Iscen, [0096] the training process can take the distillation loss into account to train the teacher model (first classification model)).
Regarding claim 20 the combination of Iscen and Zhuo teaches; The non-transitory computer-readable storage medium of claim 19, wherein determining a target loss based on the first loss and the second loss comprises: determining weight information based on a confidence for the second loss classification information (Iscen, [0087] class confidence scores are output as zi, [0088]-[0089] this function zi is a parameter in one of the loss functions which would weight/scale the parameters accordingly);
and determining a weighted sum of the first loss and the second loss according to the weight information (Iscen, [0094] the two loss terms are balanced where weight T can be adjusted to balance the weighted sum shown in the equation).
PNG
media_image8.png
226
316
media_image8.png
Greyscale
(Iscen, [0094])
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Jin (US 20160379132 A1), which pertains to learning features of social media content to further customize content provided to a user.
Ziccardi (US 20240037145 A1), which pertains to analysis of content from social media posts using machine learning to parse the posts.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JORDAN M ELLIOTT whose telephone number is (703)756-5463. The examiner can normally be reached M-F 8AM-5PM ET.
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, Emily Terrell can be reached at (571) 270-3717. 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.
/J.M.E./Examiner, Art Unit 2666 /Molly Wilburn/Primary Examiner, Art Unit 2666
Prosecution Insights