Detailed Action
Claims 1-20 are presently pending.
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 Objections
Claim 17 is objected to because of the following informalities: “obtaining domain-specific training dataset …” should read “obtaining a domain-specific training dataset”. Appropriate correction is required.
Claims 18-20 depend from claim 17 and objected to at least for the same reasons.
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 11-16 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 11 recites “processing, by the computing system, the source content with a domain-specific generative model to generate …” in lines 4-5 and “processing, by the computing system, the augmentation input and the outline with a domain-specific generative model to generate …” in lines 18-19. It is unclear whether the ‘domain-specific generative model’ in line 4 and the model in line 18 are the same generative model.
For purpose of examination, the examiner interprets the claim to mean: the first domain-specific generative model and the second domain-specific generative model are the same model.
Claims 12-16 depend from claim 11 and inherit the same deficiency.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1,
Step 1: Claim 1 recites a computing system comprising processors and non-transitory computer-readable media. Therefore, it is directed to the statutory category of a machine.
2A Prong 1:
comprises a plurality of model-generated attributes; (mental process of observation and judgment – writing a news article can be done with the aid of pen and paper)
evaluating a loss function that evaluates a difference between the model-generated news article and a respective news article of the plurality of news articles, wherein the loss function evaluates semantic differences between the model-generated news article and the respective news article and evaluates factual grounding of the model-generated news article associated with details from the press release, and wherein the loss function evaluates the plurality of model-generated attributes based on the particular news article information structure and the set of particular news article stylistic characteristics; (mathematical calculation – paragraph [0066])
adjusting one or more parameters of the generative model based at least in part on the loss function. (mathematical calculation - paragraph [0071])
2A Prong 2:
A computing system for domain-specific tuning, the system comprising: one or more processors; and (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
obtaining a domain-specific training dataset, wherein the domain-specific training dataset comprises a plurality of news articles, wherein the plurality of news articles comprise one or more domain-specific attributes associated with news articles, wherein the one or more domain-specific attributes comprise a particular news article information structure and a set of particular news article stylistic characteristics, and wherein the domain-specific training dataset comprises a plurality of respective press releases associated with the plurality of news articles; (insignificant extra-solution activity of gathering statistics MPEP 2106.05(g)(iii))
processing a press release of the plurality of respective press releases with a generative model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f) – applying the model to perform the abstract idea)
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity and combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
2B:
A computing system for domain-specific tuning, the system comprising: one or more processors; and (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
obtaining a domain-specific training dataset, wherein the domain-specific training dataset comprises a plurality of news articles, wherein the plurality of news articles comprise one or more domain-specific attributes associated with news articles, wherein the one or more domain-specific attributes comprise a particular news article information structure and a set of particular news article stylistic characteristics, and wherein the domain-specific training dataset comprises a plurality of respective press releases associated with the plurality of news articles; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
processing a press release of the plurality of respective press releases with a generative model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f) – applying the model to perform the abstract idea)
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions and usage of elements that are implemented to perform the disclosed abstract idea above.
Regarding claim 2,
Step 1: A machine, as above.
2A Prong 1: Incorporates the rejection of claim 1.
2A Prong 2:
obtaining an input dataset; (insignificant extra-solution activity of gathering statistics MPEP 2106.05(g)(iii))
processing the input dataset with the generative model to generate a domain-specific model-generated output, wherein the domain-specific model-generated output comprises a model-generated news article draft; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
processing the domain-specific model-generated output to generate a model-generated outline descriptive of a summary of substantive points within the domain-specific model-generated output; and (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
providing the model-generated outline for display. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B:
obtaining an input dataset; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
processing the input dataset with the generative model to generate a domain-specific model-generated output, wherein the domain-specific model-generated output comprises a model-generated news article draft; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
processing the domain-specific model-generated output to generate a model-generated outline descriptive of a summary of substantive points within the domain-specific model-generated output; and (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
providing the model-generated outline for display. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
Regarding claim 3,
Step 1: A machine, as above.
2A Prong 1: Incorporates the rejection of claim 2.
2A Prong 2:
obtaining an augmentation input, wherein the augmentation input is descriptive of a request to augment the model-generated outline; (insignificant extra-solution activity of gathering statistics MPEP 2106.05(g)(iii))
generating an augmented outline based on the augmentation input and the domain-specific model-generated output; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
processing the augmented outline with the generative model to generate an updated model-generated output, wherein the updated model-generated output comprises an updated model-generated news article draft; and (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
providing the updated model-generated output for display. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B:
obtaining an augmentation input, wherein the augmentation input is descriptive of a request to augment the model-generated outline; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
generating an augmented outline based on the augmentation input and the domain-specific model-generated output; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
processing the augmented outline with the generative model to generate an updated model-generated output, wherein the updated model-generated output comprises an updated model-generated news article draft; and (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
providing the updated model-generated output for display. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
Regarding claim 4,
Step 1: A machine, as above.
2A Prong 1: Incorporates the rejection of claim 3.
2A Prong 2: wherein the augmentation input is descriptive of an additional topic to add to the domain-specific model-generated output, and wherein the updated model-generated output comprises an additional section associated with the additional topic. (a field of use and technological environment MPEP 2106.05(h) – language specifying that the process steps disclosed above were used to output additional sections)
2B: wherein the augmentation input is descriptive of an additional topic to add to the domain-specific model-generated output, and wherein the updated model-generated output comprises an additional section associated with the additional topic. (a field of use and technological environment MPEP 2106.05(h) – language specifying that the process steps disclosed above were used to output additional sections)
Regarding claim 5,
Step 1: A machine, as above.
2A Prong 1: Incorporates the rejection of claim 3.
2A Prong 2: wherein the augmentation input is descriptive of a change in an order structure of the domain-specific model-generated output, and wherein the updated model-generated output comprises an updated order structure. (a field of use and technological environment MPEP 2106.05(h) – language specifying that the process steps disclosed above were used to output updated order structure)
2B: wherein the augmentation input is descriptive of a change in an order structure of the domain-specific model-generated output, and wherein the updated model-generated output comprises an updated order structure. (a field of use and technological environment MPEP 2106.05(h) – language specifying that the process steps disclosed above were used to output updated order structure)
Regarding claim 6,
Step 1: A machine, as above.
2A Prong 1: Incorporates the rejection of claim 2.
2A Prong 2: wherein processing the domain-specific model-generated output to generate the model-generated outline descriptive of the summary of substantive points within the domain-specific model-generated output comprises processing the domain-specific model-generated output with the generative model. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B: wherein processing the domain-specific model-generated output to generate the model-generated outline descriptive of the summary of substantive points within the domain-specific model-generated output comprises processing the domain-specific model-generated output with the generative model. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
Regarding claim 7,
Step 1: A machine, as above.
2A Prong 1:
evaluating a second loss function that evaluates a difference between the additional model-generated content item and one or more of the plurality of publisher content item examples; (mathematical calculation – paragraph [0071])
adjusting parameters of the generative model based at least in part on the second loss function. (mathematical calculation – paragraph [0071])
2A Prong 2:
The system of claim 1, wherein the operations further comprise: obtaining a publisher-specific dataset, wherein the publisher-specific dataset comprises a plurality of publisher content item examples; (insignificant extra-solution activity MPEP 2106.05(g)(iii) of gathering statistics)
generating an additional model-generated content item with the generative model, wherein the additional model-generated content item comprises one or more attribute features; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B:
The system of claim 1, wherein the operations further comprise: obtaining a publisher-specific dataset, wherein the publisher-specific dataset comprises a plurality of publisher content item examples; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
generating an additional model-generated content item with the generative model, wherein the additional model-generated content item comprises one or more attribute features; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
Regarding claim 8,
Step 1: A machine, as above.
2A Prong 1:
The system of claim 7, wherein evaluating the second loss function that evaluates the difference between the additional model-generated content item and the one or more of the plurality of publisher content item examples comprises: comparing the one or more attribute features of the additional model-generated content item and one or more ground truth features of the one or more of the plurality of publisher content item examples. (mathematical calculation – See paragraph [0071])
2A Prong 2: This judicial exception is not integrated into a practical application.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Regarding claim 9,
Step 1: A machine, as above.
2A Prong 1: Incorporates the rejection of claim 8.
2A Prong 2: wherein the one or more ground truth features comprise stylistic attributes associated with a publisher-specific style. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the one or more ground truth features comprise stylistic attributes associated with a publisher-specific style. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 10,
Step 1: A machine, as above.
2A Prong 1: Incorporates the rejection of claim 8.
2A Prong 2: wherein the one or more ground truth features comprise terminology attributes associated with a publisher-specific vocabulary. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the one or more ground truth features comprise terminology attributes associated with a publisher-specific vocabulary. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 11,
Step 1: Claim 11 recites a computer-implemented method comprising: obtaining, processing contents, and processing the contents to generate an outline of the content item. Therefore, it is directed to the statutory category of Processes.
2A Prong 1:
processing,
processing
processing
2A Prong 2:
A computer-implemented method, the method comprising: obtaining, by a computing system comprising one or more processors, source content, wherein the source content comprises details associated with a particular topic; (an insignificant extra-solution activity MPEP 2106.05(g)(iii))
processing, by the computing system, the source content with a domain-specific generative model … wherein the domain-specific generative model was tuned on a domain-specific training dataset to generate content items that comprise a particular information structure and a particular set of stylistic characteristics associated with news articles, and wherein the model-generated content item comprises a model-generated news article comprising one or more domain-specific attributes, wherein the one or more domain-specific attributes comprise the particular information structure and the particular set of stylistic characteristics; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
processing, by the computing system, the model-generated content item (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
providing, by the computing system, the outline of the model-generated content item for display; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
obtaining, by the computing system, an augmentation input, wherein the augmentation input is associated with augmenting the outline; and (an insignificant extra-solution activity MPEP 2106.05(g)(iii))
processing, by the computing system, the augmentation input and the outline with a domain-specific generative model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity and combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
2B:
A computer-implemented method, the method comprising: obtaining, by a computing system comprising one or more processors, source content, wherein the source content comprises details associated with a particular topic; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
processing, by the computing system, the source content with a domain-specific generative model … wherein the domain-specific generative model was tuned on a domain-specific training dataset to generate content items that comprise a particular information structure and a particular set of stylistic characteristics associated with news articles, and wherein the model-generated content item comprises a model-generated news article comprising one or more domain-specific attributes, wherein the one or more domain-specific attributes comprise the particular information structure and the particular set of stylistic characteristics; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
processing, by the computing system, the model-generated content item (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
providing, by the computing system, the outline of the model-generated content item for display; (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
obtaining, by the computing system, an augmentation input, wherein the augmentation input is associated with augmenting the outline; and (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
processing, by the computing system, the augmentation input and the outline with a domain-specific generative model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions and usage of elements that are implemented to perform the disclosed abstract idea above.
Regarding claim 12,
Step 1: Processes, as above.
2A Prong 1: Incorporates the rejection of claim 11.
2A Prong 2: further comprising: providing, by the computing system, the updated model-generated content item for display. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B: further comprising: providing, by the computing system, the updated model-generated content item for display. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
Regarding claim 13,
Step 1: Processes, as above.
2A Prong 1: Incorporates the rejection of claim 11.
2A Prong 2: wherein the source content comprises a press release and one or more interview transcripts, and wherein the model-generated content item and the updated model-generated content item are associated with the particular topic of the press release and one or more interview transcripts. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the source content comprises a press release and one or more interview transcripts, and wherein the model-generated content item and the updated model-generated content item are associated with the particular topic of the press release and one or more interview transcripts. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 14,
Step 1: Processes, as above.
2A Prong 1: Incorporates the rejection of claim 11.
2A Prong 2: wherein the outline is provided for display within a graphical user interface, (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f)) and wherein the augmentation input is received via the graphical user interface. (insignificant extra-solution activity MPEP 2106.05(g)(iii) of gathering statistics)
2B: wherein the outline is provided for display within a graphical user interface, (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f)) and wherein the augmentation input is received via the graphical user interface. (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
Regarding claim 15,
Step 1: Processes, as above.
2A Prong 1: Incorporates the rejection of claim 11.
2A Prong 2: wherein the domain-specific generative model was further tuned on a publisher-specific training dataset to generate content items that emulate a style of a particular publisher. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f) – conventional machine learning model training process)
2B: wherein the domain-specific generative model was further tuned on a publisher-specific training dataset to generate content items that emulate a style of a particular publisher. (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f) – conventional machine learning model training process)
Regarding claim 16,
Step 1: Processes, as above.
2A Prong 1: Incorporates the rejection of claim 11.
2A Prong 2: wherein the one or more domain-specific attributes comprise at least one of a domain-specific structure, a domain-specific vocabulary, or a domain-specific tone. (a field of use and technological environment MPEP 2106.05(h))
2B: wherein the one or more domain-specific attributes comprise at least one of a domain-specific structure, a domain-specific vocabulary, or a domain-specific tone. (a field of use and technological environment MPEP 2106.05(h))
Regarding claim 17,
Step 1: Claim 17 recites one or more non-transitory computer-readable media that store instructions. Therefore, it is directed to the statutory category of a machine.
2A Prong 1:
processing a particular press release of the plurality of press releases
evaluating a loss function that evaluates a difference between the model-generated article and a particular news article of respective news articles and evaluates factual grounding of the model-generated article associated with details from the particular press release, and wherein the loss function evaluates a style and structure of the model-generated article based on a comparison with a ground truth style and structure of the particular news article; (mathematical calculation – See paragraph [0067])
adjusting one or more parameters of the generative model based at least in part on the loss function. (mathematical calculation – paragraph [0071])
2A Prong 2:
One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
obtaining domain-specific training dataset, wherein the domain-specific training dataset comprises a plurality of press releases and a plurality of respective news articles, wherein the plurality of respective news articles comprise a journalistic style associated with a press style book and an inverted pyramid information structure, and wherein the plurality of respective news articles are associated with a plurality of news topics associated with the plurality of press releases; (an insignificant extra-solution activity MPEP 2106.05(g)(iii) of gathering statistics)
processing a particular press release of the plurality of press releases with a generative model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity and combination of generic computer functions that are implemented to perform the disclosed abstract idea above.
2B:
One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising: (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
obtaining domain-specific training dataset, wherein the domain-specific training dataset comprises a plurality of press releases and a plurality of respective news articles, wherein the plurality of respective news articles comprise a journalistic style associated with a press style book and an inverted pyramid information structure, and wherein the plurality of respective news articles are associated with a plurality of news topics associated with the plurality of press releases; (indicated as an insignificant extra-solution activity MPEP 2106.05(g)(iii) in Step 2A Prong 2. Therefore, it is re-evaluated in Step 2B as well understood, routine and conventional activity MPEP 2106.05(d)(II)(iv) of gathering statistics)
processing a particular press release of the plurality of press releases with a generative model (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are well, understood, routine and conventional activity as disclosed in combination of generic computer functions and usage of elements that are implemented to perform the disclosed abstract idea above.
Regarding claim 18,
Step 1: A machine, as above.
2A Prong 1: wherein the loss function further evaluates the model-generated article based on a structural comparison between content of the model-generated article and the particular news article of respective news articles. (mathematical calculation – See paragraph [0067])
2A Prong 2: The one or more non-transitory computer-readable media of claim 17 (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B: The one or more non-transitory computer-readable media of claim 17 (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
Regarding claim 19,
Step 1: A machine, as above.
2A Prong 1: wherein the loss function further evaluates the model-generated article based on a verbatim penalization term, wherein the verbatim penalization term adjusts a gradient descent based on a verbatim similarity measure between the model-generated article and at least one of the particular news article or the particular press release. (mathematical calculation – See paragraph [0067])
2A Prong 2: The one or more non-transitory computer-readable media of claim 17 (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B: The one or more non-transitory computer-readable media of claim 17 (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
Regarding claim 20,
Step 1: A machine, as above.
2A Prong 1: wherein the loss function further evaluates the model-generated article based on an attribution penalization term, wherein the attribution penalization term adjusts a gradient descent based on evaluating a quality of an attribution within the model-generated article. (mathematical calculation – See paragraph [0067])
2A Prong 2: The one or more non-transitory computer-readable media of claim 17 (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
2B: The one or more non-transitory computer-readable media of claim 17 (mere instructions to apply an exception using a generic computer component MPEP 2106.05(f))
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.
Claim 1 is rejected under 35 U.S.C. 103 as being unpatentable over Mosallanezhad Mosallanezhad et al., (“Generating Topic-Preserving Synthetic News”, 2021, hereinafter ‘Mosallanezhad’) in view of Andersen et al. (“Automatic Extraction of Facts from Press Releases to Generate News Stories”, 1992, hereinafter ‘Andersen’).
Regarding claim 1, Mosallanezhad teaches:
A computing system for domain-specific tuning, the system comprising: one or more processors; and ([Mosallanezhad, page 496, left col, lines 16-20] shows that the model focuses on domain-specific news generation. [Mosallanezhad, page 495, left col, 2nd para, lines 7-14] discloses that the training process involves construction of a memory array, which indicates that the step is performed in a generic computer)
one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising: ([Mosallanezhad, page 495, left col, 2nd para, lines 7-14] discloses that the training process involves construction of a memory array, which indicates that the step is performed in a generic computer)
obtaining a domain-specific training dataset, wherein the domain-specific training dataset comprises a plurality of news articles, wherein the plurality of news articles comprise one or more domain-specific attributes associated with news articles, wherein the one or more domain-specific attributes comprise a particular news article information structure and a set of particular news article stylistic characteristics, and wherein the domain-specific training dataset [Mosallanezhad, page 492, left col, Problem Statement], [page 493, right col, B. Using Adversaries to Generate Realistic Synthetic News] and [page 494, right col, A. Data, lines 1-4] collectively discloses utilizing news data X from two different platforms. For training the agent, news dataset
X
=
{
S
0
1
,
x
1
,
…
,
S
0
N
,
x
N
}
in which
S
0
N
shows the topic of
i
t
h
news (i.e., press release) and
x
i
shows the content of that news (i.e., news articles). [page 493, left col, last para – right col, lines 5] discloses that the
X
includes news article stylistic characteristics and the agent is trained to copy the writing style of the news article. The ‘similar word sequence’ is the article information structure)
processing a [Mosallanezhad, page 493, left col, lines 2-17] and [page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions, and the reward function (i.e., loss function) evaluates similarity between the generated news contents and the reference news contents. This paragraph indicates that the input data X which includes topics and news contents is processed by the agent (generative model) to generate synthetic news articles prior to the training process)
evaluating a loss function that evaluates a difference between the model-generated news article and a respective news article of the plurality of news articles, wherein the loss function evaluates semantic differences between the model-generated news article and the respective news article and evaluates factual grounding of the model-generated news article associated with details from the press release, and wherein the loss function evaluates the plurality of model-generated attributes based on the particular news article information structure and the set of particular news article stylistic characteristics; and ([page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights. The loss function is calculated based on loss function
L
(
θ
)
which includes the reward function
r
t
+
1
that calculates the similarities. [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function
r
t
to learn the strategy for selecting actions. The reward function
r
t
evaluates similarity between the generated news contents and the reference news contents. Cosine Similarity measures similarities between generated news topic
S
t
and reference news topic
S
0
(i.e., factual grounding associated with details from the news topic). BLEU score measures how many words overlap between the generated news
S
t
and the reference news contents
X
(i.e., semantic difference) to maintain the writing style of news content (i.e., information structure and stylistic characteristics) )
adjusting one or more parameters of the generative model based at least in part on the loss function. ([Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions. [page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights)
However, Mosallanezhad does not specifically disclose:
dataset comprises a plurality of respective press releases associated with the plurality of news articles
processing a press release.
Andersen teaches:
dataset comprises a plurality of respective press releases associated with the plurality of news articles ([Andersen, page 172, right col, 4. Technical Approach, lines 1-25] discloses receiving a press release from PR Newswire, and then extracting relevant information from the press release to generate news stories)
processing a press release ([Andersen, page 172, right col, 4. Technical Approach, lines 1-25] discloses receiving a press release from PR Newswire, and then extracting relevant information from the press release to generate news stories)
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of receiving dataset comprises a plurality of press releases of Andersen to alter the ‘topic (first few words)’ of Mosallanezhad to implement the method of the present invention. The suggestion and/or motivation for doing so is to generate news articles with more accurate information as press releases are generally used as a primary source for information by reporters.
Claims 2 and 6 are rejected under 35 U.S.C. 103 as being unpatentable over Mosallanezhad in view of Andersen and further in view of Saito et al. (US 20220343076 A1, hereinafter ‘Saito’).
Regarding claim 2, Mosallanezhad teaches:
The system of claim 1, wherein the operations further comprise: obtaining an input dataset; ([Mosallanezhad, page 492, left col, Problem Statement], [page 493, right col, B. Using Adversaries to Generate Realistic Synthetic News] and [page 494, right col, A. Data, lines 1-4] collectively discloses utilizing news data X from two different platforms. For training the agent, news dataset
X
=
{
S
0
1
,
x
1
,
…
,
S
0
N
,
x
N
}
in which
S
0
N
shows the topic of
i
t
h
news (i.e., press release) and
x
i
shows the content of that news (i.e., news articles). [page 493, left col, last para – right col, lines 5] discloses that the
X
includes news article stylistic characteristics and the agent is trained to copy the writing style of the news article)
processing the input dataset with the generative model to generate a domain-specific model-generated output, wherein the domain-specific model-generated output comprises a model-generated news article draft; ([Mosallanezhad, page 493, left col, lines 2-17] and [page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions, and the reward function (i.e., loss function) evaluates similarity between the generated news contents and the reference news contents. This paragraph indicates that the input data X which includes topics and news contents is processed by the agent (generative model) to generate synthetic news articles prior to the training process)
However, Mosallanezhad in view of Andersen do not specifically disclose:
processing the domain-specific model-generated output to generate a model-generated outline descriptive of a summary of substantive points within the domain-specific model-generated output; and
providing the model-generated outline for display.
Saito teaches:
processing the domain-specific model-generated output to generate a model-generated outline descriptive of a summary of substantive points within the domain-specific model-generated output; and ([Saito, Fig. 8] and [0042-0045] Source text encoding unit 121 and [0046-0047] reference text encoding unit 122 processes source text and reference text, respectively. [0049-0052] Decoding unit 123 receives the outputs from 121 and 122 to generate outputs (i.e., model-generated output) for Synthesis Unit 124. [0053-0054] Synthesis Unit 124 generates summary (i.e., model-generated outline) of the source text)
providing the model-generated outline for display. ([Saito, 0053-0054] Synthesis Unit 124 generates summary (i.e., model-generated outline) of the source text. [0071] shows that summaries are bullets for articles displayed on websites)
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the outline generation and display method of Saito to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to improve the efficiency of the news article generation method by reducing the effort required by users to summarize the generated news articles.
Regarding claim 6, Mosallanezhad in view of Andersen in view of Saito teaches:
The system of claim 2, wherein processing the domain-specific model-generated output to generate the model-generated outline descriptive of the summary of substantive points within the domain-specific model-generated output comprises processing the domain-specific model-generated output with the generative model. ([Saito, Fig. 8] and [0042-0045] Source text encoding unit 121 and [0046-0047] reference text encoding unit 122 processes source text and reference text, respectively. [0049-0052] Decoding unit 123 receives the outputs from 121 and 122 to generate outputs (i.e., model-generated output) for Synthesis Unit 124. [0053-0054] Synthesis Unit 124 generates summary (i.e., model-generated outline) of the source text)
Claims 3-5 are rejected under 35 U.S.C. 103 as being unpatentable over Masallanezhad in view of Andersen in view of Saito and further in view of Tsourma et al. (“An AI-Enabled Framework for Real-Time Generation of News Articles Based on Big EO Data for Disaster Reporting”, 2021, hereinafter ‘Tsourma’).
Regarding claim 3, Masallanezhad in view of Andersen and further in view of Saito teaches:
The system of claim 2.
However, Masallanezhad in view of Andersen and further in view of Saito do not disclose:
wherein the operations further comprise:
obtaining an augmentation input, wherein the augmentation input is descriptive of a request to augment the model-generated outline;
generating an augmented outline based on the augmentation input and the domain-specific model-generated output;
processing the augmented outline with the generative model to generate an updated model-generated output, wherein the updated model-generated output comprises an updated model-generated news article draft; and
providing the updated model-generated output for display.
Tsourma teaches:
wherein the operations further comprise: obtaining an augmentation input, wherein the augmentation input is descriptive of a request to augment the model-generated outline; ([Tsourma, page 11, 2.10.3. News Article Text Synthesis, lines 1-17] and [page 13, last para, line 1 – page 14, line 12] collectively disclose that the user can select from available topics (domains), articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
generating an augmented outline based on the augmentation input and the domain-specific model-generated output; ([Tsourma, page 10, 2.10. EarthBot, lines 8-21] EarthBot generates news articles based on input social media and news articles relevant to a disaster (domain) and also receives the journalist’s profile as its input for the personalization of the generated text. [page 13, last para, line 1 – page 14, line 12] discloses that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
processing the augmented outline with the generative model to generate an updated model-generated output, wherein the updated model-generated output comprises an updated model-generated news article draft; and ([Tsourma, page 13, last para, line 1 – page 14, line 12] discloses that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
providing the updated model-generated output for display. ([Tsourma, page 13, last para, line 1 – page 14, line 12] discloses that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output))
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of augmenting additional data to the outline and model-generated output of Tsourma to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to improve the efficiency of the news article generation method by reducing the effort required by users to re-write the full summary and the news article when new information is introduced.
Regarding claim 4, Mosallanezhad in view of Andersen in view of Saito and further in view of Tsourma teach:
The system of claim 3, wherein the augmentation input is descriptive of an additional topic to add to the domain-specific model-generated output, and wherein the updated model-generated output comprises an additional section associated with the additional topic. ([Tsourma, page 11, 2.10.3. News Article Text Synthesis, lines 1-17] and [page 13, last para, line 1 – page 14, line 12] collectively disclose that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
Regarding claim 5, Mosallanezhad in view of Andersen in view of Saito and further in view of Tsourma teach:
The system of claim 3, wherein the augmentation input is descriptive of a change in an order structure of the domain-specific model-generated output, and wherein the updated model-generated output comprises an updated order structure. ([Tsourma, page 11, 2.10.3. News Article Text Synthesis, lines 1-17] and [page 13, last para, line 1 – page 14, line 12] collectively disclose that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords (order structure), remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
Claims 7-10 are rejected under 35 U.S.C. 103 as being unpatentable over Mosallanezhad in view of Andersen and further in view of McAteer et al. (US 20200004875 A1, hereinafter ‘McAteer’).
Regarding claim 7, Mosallanezhad teaches:
The system of claim 1, wherein the operations further comprise: obtaining a [Mosallanezhad, page 492, left col, Problem Statement], [page 493, right col, B. Using Adversaries to Generate Realistic Synthetic News] and [page 494, right col, A. Data, lines 1-4] collectively discloses utilizing news dataset X from two different platforms. For training the agent, news dataset
X
=
{
S
0
1
,
x
1
,
…
,
S
0
N
,
x
N
}
in which
S
0
N
shows the topic of
i
t
h
news (i.e., press release) and
x
i
shows the content of that news (i.e., news articles).(
S
0
2
,
x
N
)
which is the second news article is the additional content. [page 493, left col, last para – right col, lines 5] discloses that the
X
includes news article stylistic characteristics and the agent is trained to copy the writing style of the news article)
generating an additional model-generated content item with the generative model, wherein the additional model-generated content item comprises one or more attribute features; ([Mosallanezhad, page 493, left col, lines 2-17] and [page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions, and the reward function (i.e., loss function) evaluates similarity between the generated news contents and the reference news contents. This paragraph indicates that the input data X which includes topics and news contents is processed by the agent (generative model) to generate synthetic news articles prior to the training process)
evaluating a second loss function that evaluates a difference between the additional model-generated content item and one or more of the plurality of [page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights. The loss function is calculated based on loss function
L
(
θ
)
which includes the reward function
r
t
+
1
that calculates the similarities. [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function
r
t
to learn the strategy for selecting actions. The reward function
r
t
evaluates similarity between the generated news contents and the reference news contents. Cosine Similarity measures similarities between generated news topic
S
t
and reference news topic
S
0
(i.e., factual grounding associated with details from the news topic). BLEU score measures how many words overlap between the generated news
S
t
and the reference news contents
X
(i.e., semantic difference) to maintain the writing style of news content (i.e., information structure and stylistic characteristics) )
adjusting parameters of the generative model based at least in part on the second loss function. ([Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions. [page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights)
However, Mosallanezhad and Andersen do not specifically disclose:
obtaining a publisher specific dataset
McAteer teaches:
obtaining a publisher specific dataset ([McAteer, 0070] The ground truth for a domain includes common types of phrases and vocabulary representative of typical users (publisher))
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of obtaining a publisher specific dataset of McAteer to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to utilize more diverse information from the reference articles in order to generate synthesized news articles that more closely resemble the reference articles.
Regarding claim 8, Mosallanezhad teaches:
The system of claim 7, wherein evaluating the second loss function that evaluates the difference between the additional model-generated content item and the one or more of the plurality of [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions. The reward function (i.e., loss function) evaluates similarity between the generated news contents and the reference news contents. Cosine Similarity measures similarities between generated news topic
S
t
and reference news topic
S
0
(i.e., factual grounding associated with details from the news topic). [page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights)
However, Mosallanezhad and Andersen do not specifically disclose:
the publisher content item
McAteer teaches:
the publisher content item ([McAteer, 0070] The ground truth for a domain includes common types of phrases and vocabulary representative of typical users (publisher))
Regarding claim 9, Mosallanezhad teaches:
The system of claim 8, wherein the one or more ground truth features comprise stylistic attributes associated [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions. The reward function (i.e., loss function) evaluates similarity between the generated news contents and the reference news contents. BLEU score measures how many words overlap between the generated news
S
t
and the reference news contents
X
(i.e., semantic difference) to maintain the writing style of news content (i.e., information structure and stylistic characteristics). [page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights)
However, Mosallanezhad and Andersen do not specifically disclose:
wherein the one or more ground truth features comprise stylistic attributes associated with a publisher-specific style
McAteer teaches:
wherein the one or more ground truth features comprise stylistic attributes associated with a publisher-specific style ([McAteer, 0070] The ground truth for a domain includes common types of phrases and vocabulary representative of typical users (publisher))
Regarding claim 10, Mosallanezhad in view of Andersen and further in view of McAteer teaches:
The system of claim 8, wherein the one or more ground truth features comprise terminology attributes associated with a publisher-specific vocabulary. ([McAteer, 0070] The ground truth for a domain includes common types of phrases and vocabulary representative of typical users (publisher))
Claims 11-12, 14 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Masallanezhad in view of Saito and further in view of Tsourma.
Regarding claim 11, Mosallanezhad teaches:
A computer-implemented method, the method comprising: obtaining, by a computing system comprising one or more processors, source content, wherein the source content comprises details associated with a particular topic; ([Mosallanezhad, page 496, left col, lines 16-20] shows that the model focuses on domain-specific news generation. [Mosallanezhad, page 495, left col, 2nd para, lines 7-14] discloses that the training process involves construction of a memory array, which indicates that the step is performed in a generic computer. [Mosallanezhad, page 493, right col, B. Using Adversaries to Generate Realistic Synthetic News] and [page 494, right col, A. Data, lines 1-4] collectively discloses utilizing news data X (i.e., source content) from two different platforms. For training the agent, news dataset
X
=
{
S
0
1
,
x
1
,
…
,
S
0
N
,
x
N
}
in which
S
0
N
shows the topic of
i
t
h
news (i.e., press release) and
x
i
shows the content of that news (i.e., news articles))
processing, by the computing system, the source content with a domain-specific generative model to generate a model-generated content item, wherein the domain-specific generative model was tuned on a domain-specific training dataset to generate content items that comprise a particular information structure and a particular set of stylistic characteristics associated with news articles, and wherein the model-generated content item comprises a model-generated news article comprising one or more domain-specific attributes, wherein the one or more domain-specific attributes comprise the particular information structure and the particular set of stylistic characteristics; ([page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights. The loss function is calculated based on loss function
L
(
θ
)
which includes the reward function
r
t
+
1
that calculates the similarities. [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function
r
t
to learn the strategy for selecting actions. The reward function
r
t
evaluates similarity between the generated news contents and the reference news contents. Cosine Similarity measures similarities between generated news topic
S
t
and reference news topic
S
0
(i.e., factual grounding associated with details from the news topic). BLEU score measures how many words overlap between the generated news
S
t
and the reference news contents
X
(i.e., semantic difference) to maintain the writing style of news content (i.e., information structure and stylistic characteristics) )
However, Mosallanezhad does not specifically disclose:
processing, by the computing system, the model-generated content item to generate an outline of the model-generated content item;
providing, by the computing system, the outline of the model-generated content item for display;
obtaining, by the computing system, an augmentation input, wherein the augmentation input is associated with augmenting the outline; and
processing, by the computing system, the augmentation input and the outline with a domain-specific generative model to generate an updated model-generated content item, wherein the updated model-generated content item comprises an updated model-generated news article.
Saito teaches:
processing, by the computing system, the model-generated content item to generate an outline of the model-generated content item; ([Saito, Fig. 8] and [0042-0045] Source text encoding unit 121 and [0046-0047] reference text encoding unit 122 processes source text and reference text, respectively. [0049-0052] Decoding unit 123 receives the outputs from 121 and 122 to generate outputs (i.e., model-generated output) for Synthesis Unit 124. [0053-0054] Synthesis Unit 124 generates summary (i.e., model-generated outline) of the source text)
providing, by the computing system, the outline of the model-generated content item for display; ([Saito, 0053-0054] Synthesis Unit 124 generates summary (i.e., model-generated outline) of the source text. [0071] shows that summaries are bullets for articles displayed on websites)
However, Mosallanezhad in view of Saito do not specifically disclose:
obtaining, by the computing system, an augmentation input, wherein the augmentation input is associated with augmenting the outline; and
processing, by the computing system, the augmentation input and the outline with a domain-specific generative model to generate an updated model-generated content item, wherein the updated model-generated content item comprises an updated model-generated news article.
Tsourma teaches:
obtaining, by the computing system, an augmentation input, wherein the augmentation input is associated with augmenting the outline; and ([Tsourma, page 11, 2.10.3. News Article Text Synthesis, lines 1-17] and [page 13, last para, line 1 – page 14, line 12] collectively disclose that the user can select from the available topics (domains), articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
processing, by the computing system, the augmentation input and the outline with a domain-specific generative model to generate an updated model-generated content item, wherein the updated model-generated content item comprises an updated model-generated news article. ([Tsourma, page 10, 2.10. EarthBot, lines 8-21] EarthBot generates news articles based on input social media and news articles relevant to a disaster and also receives the journalist’s profile as its input for the personalization of the generated text. [page 13, last para, line 1 – page 14, line 12] discloses that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of augmenting additional data to the outline and model-generated output of Tsourma to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to improve the efficiency of the news article generation method by reducing the effort required by users to re-write the full summary and the news article when new information is introduced.
Regarding claim 12, Mosallanezhad in view of Saito and further in view of Tsourma teach:
The method of claim 11, further comprising: providing, by the computing system, the updated model-generated content item for display. ([Tsourma, page 13, last para, line 1 – page 14, line 12] discloses that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output))
Regarding claim 14, Mosallanezhad in view of Saito teach:
The method of claim 11, wherein the outline is provided for display within a graphical user interface, ([Saito, 0053-0054] Synthesis Unit 124 generates summary (i.e., model-generated outline) of the source text. [0071] shows that summaries are bullets for articles displayed on websites)
However, Mosallanezhad in view of Saito do not specifically teach:
and wherein the augmentation input is received via the graphical user interface.
Tsourma teaches:
and wherein the augmentation input is received via the graphical user interface. ([Tsourma, page 11, 2.10.3. News Article Text Synthesis, lines 1-17] and [page 13, last para, line 1 – page 14, line 12] collectively disclose that the user can select articles, posts, or images for the generation of the final article, and the user also can re-arrange the order of keywords, remove the keywords, or include new keywords and re-generate a text and view the newly generated article (updated model-generated output). The user selection of edit method is the request to augment the model-generated article)
Regarding claim 16, Mosallanezhad teaches:
The method of claim 11, wherein the one or more domain-specific attributes comprise at least one of a domain-specific structure, a domain-specific vocabulary, or a domain-specific tone. ([page 493, right col, last para, lines 1-4] One of the attributes the model should consider is the word sequence (domain-specific structure) of the reference news and the generated news)
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Mosallanezhad in view of Saito in view of Tsourma in view of Andersen and further in view of SMITH LEWIS et al. (US 20240289863 A1, hereinafter ‘SMITH LEWIS’).
Regarding claim 13, Mosallanezhad teaches:
The method of claim 11, wherein the source content comprises a press release generated content item are associated with the particular topic [Mosallanezhad, page 493, left col, lines 2-17] and [page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions, and the reward function (i.e., loss function) evaluates similarity between the generated news contents and the reference news contents. This paragraph indicates that the input data X which includes topics and news contents is processed by the agent (generative model) to generate synthetic news articles prior to the training process)
However, Mosallanezhad in view of Saito and further in view of Moriarty does not specifically disclose:
wherein the source content comprises a press release and one or more interview transcripts.
Andersen teaches:
wherein the source content comprises a press release ([Andersen, page 172, right col, 4. Technical Approach, lines 1-25] discloses receiving a press release from PR Newswire, and then extracting relevant information from the press release to generate news stories)
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of receiving dataset comprises a plurality of press releases of Andersen to alter the ‘topic (first few words)’ of Mosallanezhad to implement the method of the present invention. The suggestion and/or motivation for doing so is to generate news articles with more accurate information as press releases are generally used as a primary source for information by reporters.
However, Mosallanezhad in view of Saito in view of Tsourma and further in view of Andersen do not specifically disclose:
wherein the source content comprises one or more interview transcripts
SMITH LEWIS teaches:
wherein the source content comprises one or more interview transcripts ([0111] discloses using interview transcripts, news articles, and policy documents to train an AI model)
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of obtaining both press release and interview transcripts of SMITH LEWIS to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to utilize more diverse information from the reference articles in order to generate synthesized news articles that more closely resemble the reference articles.
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Mosallanezhad in view of Saito in view of Tsourma and further in view of McAteer.
Regarding claim 15, Mosallanezhad in view of Saito and further in view of Tsourma teaches:
The method of claim 11.
However, Mosallanezhad in view of Saito and further in view of Tsourma does not specifically disclose:
wherein the domain-specific generative model was further tuned on a publisher-specific training dataset to generate content items that emulate a style of a particular publisher.
McAteer teaches:
wherein the domain-specific generative model was further tuned on a publisher-specific training dataset to generate content items that emulate a style of a particular publisher. ([McAteer, 0070] The ground truth for a domain includes common types of phrases and vocabulary representative of typical users (publisher))
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of obtaining a publisher specific dataset of McAteer to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to utilize more diverse information from the reference articles in order to generate synthesized news articles that more closely resemble the reference articles.
Claim 17-19 are rejected under 35 U.S.C. 103 as being unpatentable over Mosallanezhad in view of Andersen and further in view of Liu et al. (“Conditional Neural Generation using Sub-Aspect Functions for Extractive News Summarization”, 2020, hereinafter ‘Liu’).
Regarding claim 17, Mosallanezhad teaches:
One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising: ([Mosallanezhad, page 496, left col, lines 16-20] shows that the model focuses on domain-specific news generation. [Mosallanezhad, page 495, left col, 2nd para, lines 7-14] discloses that the training process involves construction of a memory array, which indicates that the step is performed in a generic computer)
obtaining domain-specific training dataset, wherein the domain-specific training dataset comprises a plurality of [Mosallanezhad, page 493, right col, B. Using Adversaries to Generate Realistic Synthetic News] and [page 494, right col, A. Data, lines 1-4] collectively discloses utilizing news data X from two different platforms. For training the agent, news dataset
X
=
{
S
0
1
,
x
1
,
…
,
S
0
N
,
x
N
}
in which
S
0
N
shows the topic of
i
t
h
news (i.e., press release) and
x
i
shows the content of that news (i.e., news articles))
processing a particular [Mosallanezhad, page 493, left col, lines 2-17] and [page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions, and the reward function (i.e., loss function) evaluates similarity between the generated news contents and the reference news contents. This paragraph indicates that the input data X which includes topics and news contents is processed by the agent (generative model) to generate synthetic news articles prior to the training process)
evaluating a loss function that evaluates a difference between the model-generated article and a particular news article of respective news articles and evaluates factual grounding of the model-generated article associated with details from the particular press release, and wherein the loss function evaluates a style and structure of the model-generated article based on a comparison with a ground truth style and structure of the particular news article; and ([page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights. The loss function is calculated based on loss function
L
(
θ
)
which includes the reward function
r
t
+
1
that calculates the similarities. [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function
r
t
to learn the strategy for selecting actions. The reward function
r
t
evaluates similarity between the generated news contents and the reference news contents. Cosine Similarity measures similarities between generated news topic
S
t
and reference news topic
S
0
(i.e., factual grounding associated with details from the news topic). BLEU score measures how many words overlap between the generated news
S
t
and the reference news contents
X
(i.e., semantic difference) to maintain the writing style of news content (i.e., information structure and stylistic characteristics) )
adjusting one or more parameters of the generative model based at least in part on the loss function. ([Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function to learn the strategy for selecting actions. [page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights)
However, Mosallanezhad does not specifically disclose:
dataset comprises a plurality of press releases
processing press releases
news articles comprise a journalistic style associated with a press style book and an inverted pyramid information structure
Andersen teaches:
dataset comprises a plurality of press releases ([Andersen, page 172, right col, 4. Technical Approach, lines 1-25] discloses receiving a press release from PR Newswire, and then extracting relevant information from the press release to generate news stories)
processing press releases ([Andersen, page 172, right col, 4. Technical Approach, lines 1-25] discloses receiving a press release from PR Newswire, and then extracting relevant information from the press release to generate news stories)
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of receiving dataset comprises a plurality of press releases of Andersen to alter the ‘topic (first few words)’ of Mosallanezhad to implement the method of the present invention. The suggestion and/or motivation for doing so is to generate news articles with more accurate information as press releases are generally used as a primary source for information by reporters.
However, Mosallanezhad in view of Andersen do not specifically disclose:
news articles comprise a journalistic style associated with a press style book and an inverted pyramid information structure
Liu teaches:
news articles comprise a journalistic style associated with a press style book and an inverted pyramid information structure ([Liu, page 1456, right col, lines 5-17] and [page 1458, left col, lines 1-6] collectively disclose adopting sub-function features including (1) inverted pyramid writing style is common, (2) Importance sub-aspect indicates the assumption that repeatedly occurring content in the source document contains more important information (3) Diversity sub-aspect suggest that selected salient sentences should maximize the semantic volume in a distributed semantic space, to generate news summarizations)
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the method of using news articles comprising a journalistic style associated with a press style book and an inverted pyramid information structure of Liu to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to utilize more diverse format information from the reference articles in order to generate synthesized news articles that more closely resemble the reference news articles.
Regarding claim 18, Mosallanezhad teaches:
The one or more non-transitory computer-readable media of claim 17, wherein the loss function further evaluates the model-generated article based on a structural comparison between content of the model-generated article and the particular news article of respective news articles. ([page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights. The loss function is calculated based on loss function
L
(
θ
)
which includes the reward function
r
t
+
1
that calculates the similarities. [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function
r
t
to learn the strategy for selecting actions. The reward function
r
t
evaluates similarity between the generated news contents and the reference news contents. Cosine Similarity measures similarities between generated news topic
S
t
and reference news topic
S
0
(i.e., factual grounding associated with details from the news topic). BLEU score measures how many words overlap between the generated news
S
t
and the reference news contents
X
(i.e., semantic difference) to maintain the writing style of news content (i.e., information structure and stylistic characteristics) )
Regarding claim 19, Mosallanezhad teaches:
wherein the loss function further evaluates the model-generated article based on a verbatim penalization term, wherein the verbatim penalization term adjusts a gradient descent based on a verbatim similarity measure between the model-generated article and at least one of the particular news article or the particular press release. ([page 494, right col, Algorithm 1, lines 10-11] The learning process involves update of DQN weights. The loss function is calculated based on loss function
L
(
θ
)
which includes the reward function
r
t
+
1
that calculates the similarities. [Mosallanezhad, page 493, left col, Reward Function, line 1 – right col, line 9] The agent uses a reward function
r
t
to learn the strategy for selecting actions. The reward function
r
t
evaluates similarity between the generated news contents and the reference news contents. Cosine Similarity measures similarities between generated news topic
S
t
and reference news topic
S
0
(i.e., factual grounding associated with details from the news topic). BLEU score measures how many words overlap between the generated news
S
t
and the reference news contents
X
(i.e., semantic difference) to maintain the writing style of news content (i.e., information structure and stylistic characteristics) )
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Mosallanezhad in view of Andersen in view of Liu and further in view of Helbling et al. (US 20240412420 A1, hereinafter ‘Helbling’).
Regarding claim 20, Mosallanezhad in view of Andersen and further in view of Liu teaches:
The one or more non-transitory computer-readable media of claim 17.
However, Mosallanezhad in view of Andersen and further in view of Liu do not specifically disclose:
wherein the loss function further evaluates the model-generated article based on an attribution penalization term, wherein the attribution penalization term adjusts a gradient descent based on evaluating a quality of an attribution within the model-generated article.
Helbling teaches:
wherein the loss function further evaluates the model-generated article based on an attribution penalization term, wherein the attribution penalization term adjusts a gradient descent based on evaluating a quality of an attribution within the model-generated article. ([Helbling, 0098]-[0099] discloses using the cosine similarity between V(x) text encoder and the attribute direction as a penalty term to preserve the intensity of a relative attribute. The generative model may further modify the attribute and minimize the distance value (i.e., adjusts a gradient descent))
It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to use the attribution penalization term that adjusts a gradient descent based on evaluating a quality of Helbling to implement the news article generation method of Mosallanezhad. The suggestion and/or motivation for doing so is to generate synthesized news articles that more closely resemble the reference news articles by comparing attributions of the reference articles and the generated articles.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JUN KWON whose telephone number is (571)272-2072. The examiner can normally be reached Monday – Friday 8:00AM – 5:00PM 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, Abdullah Kawsar can be reached at (571)270-3169. 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.
/JUN KWON/Examiner, Art Unit 2127
/ABDULLAH AL KAWSAR/Supervisory Patent Examiner, Art Unit 2127
Prosecution Insights