DETAILED ACTION
This non-final office action is responsive to application 18/767,225 as submitted on July 09, 2024.
Claims 1-18 are canceled.
Claims 19-38 are added.
Claim status is currently pending and under examination for claims 19-38 of which independent claims are 19, 24, 29 and 34.
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 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 25-28, 30-33 and 35-38 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.
Dependent Claim 25 recites the limitation “the electronic infrastructure” in line 1. There is a lack of antecedent basis for “the electronic infrastructure”, therefore the claim is rendered indefinite. For examination, “the electronic infrastructure” is interpreted to refer to “an artificial intelligence infrastructure”.
Dependent Claims 26-28, 30-33 and 35-38 recite similar limitations as claim 25, therefore the same rationale of rejection is applied.
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 19-38 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Independent Claim 19
Step 2A Prong One: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, independent claim 19, under the broadest reasonable interpretation, recites the following limitations that are abstract ideas:
and processing circuitry operable to perform influence balancing and merger operations on the plurality of influence data to produce combined influence data destined to influence an operational neural network. (mental process)
The “perform” step involves weighting and combining data to create aggregate data which amounts to no more than observations, evaluations, and judgments that can be performed in the human mind or with the use of a physical aid (e.g., pen and paper). The claim recites the step of performing influence balancing and merger operations at a high degree of generality, thus the step is not required to have any specific level of complexity that would preclude the step from being mental processes. Therefore, the “perform” step is considered to be mental processes, see MPEP § 2106.04(a)(2)(III).
Therefore, the independent claim recites a judicial exception.
Step 2A Prong Two: Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, the judicial exception recited above is not integrated into a practical application. The claims recite the following additional elements, but these additional elements are not sufficient to integrate the judicial exception into a practical application:
a plurality of influence data originating from a corresponding plurality of data sources; (MPEP § 2106.05(g) necessary data gathering and insignificant extra-solution activity to the judicial exception)
and processing circuitry operable to perform influence balancing and merger operations on the plurality of influence data to produce combined influence data destined to influence an operational neural network. (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or generally links exception to a technological environment)
The “a plurality of influence data originating …” step amounts to mere data gathering and is recited at a high level of generality, thus adding insignificant extra-solution activity to the judicial exception – see MPEP § 2106.05(g). Under MPEP § 2106.05(d), such additional elements have been found by the courts to not integrate a judicial exception into a practical application.
The “perform” step requires “processing circuitry” to perform influence balancing and merger operations. The processing circuitry is used to apply the recited judicial exception without placing any limitation on how the processing circuitry operates. The limitation amounts to mere instructions to “apply” the judicial exception on a computer. It can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
Therefore, the above limitations do not integrate the judicial exception into a practical application.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception?
No. The claims do not include additional elements that are sufficient for the claims to amount to significantly more than the judicial exception.
In regards to the “a plurality of influence data originating …” step, this step adds insignificant extra-solution activity. An extra-solution activity is a well-understood, routine and conventional (WURC) activity per MPEP § 2106.05(d)(II), “the courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. i. Receiving or transmitting data over a network, e.g., using the Internet to gather data.” The step does not integrate the judicial exception into a practical application and does not amount to significantly more.
In regards to the “processing circuitry” in the “perform” step, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
Therefore, independent claim 19 is not patent eligible.
Independent Claim 24
Step 2A Prong One: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, independent claim 24, under the broadest reasonable interpretation, recites the following limitations that are abstract ideas:
and a support node being configured to respond to first data generated by the first neural network node by producing, from at least the first data, second data that is used to influence the first neural network node. (mental process)
The “respond” step involves using first data to determine second data which amounts to no more than observations, evaluations, and judgments that can be performed in the human mind or with the use of a physical aid (e.g., pen and paper). The claim recites the step of responding to first data at a high degree of generality, thus the step is not required to have any specific level of complexity that would preclude the step from being mental processes. Therefore, the “respond” step is considered to be mental processes, see MPEP § 2106.04(a)(2)(III).
Therefore, the independent claim recites a judicial exception.
Step 2A Prong Two: Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, the judicial exception recited above is not integrated into a practical application. The claims recite the following additional elements, but these additional elements are not sufficient to integrate the judicial exception into a practical application:
a first neural network node; (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea)
a second neural network node; (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea)
and a support node being configured to respond to first data generated by the first neural network node by producing, from at least the first data, second data that is used to influence the first neural network node. (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or generally links exception to a technological environment)
The “respond” step requires a “support node” to respond to first data. The support node is used to apply the recited judicial exception without placing any limitation on how the support node operates. The limitation amounts to mere instructions to “apply” the judicial exception on a computer. It can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
The remaining additional elements are recited at a high-level of generality such that they amount to no more than mere instructions to “apply” an exception using a generic component. Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, see MPEP § 2106.05(f).
Therefore, the above limitations do not integrate the judicial exception into a practical application.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception?
No. The claims do not include additional elements that are sufficient for the claims to amount to significantly more than the judicial exception.
In regards to the “support node” in the “respond” step, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
In regards to the remaining additional elements, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
Therefore, independent claim 24 is not patent eligible.
Independent Claim 29
Step 2A Prong One: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, independent claim 29, under the broadest reasonable interpretation, recites the following limitations that are abstract ideas:
a support processing node being configured to apply influence balancing to a plurality of data inputs to produce combined influence data; (mental process)
The “apply” step involves weighting data inputs to determine aggregate data which amounts to no more than observations, evaluations, and judgments that can be performed in the human mind or with the use of a physical aid (e.g., pen and paper). The claim recites the step of applying influence balancing at a high degree of generality, thus the step is not required to have any specific level of complexity that would preclude the step from being mental processes. Therefore, the “apply” step is considered to be mental processes, see MPEP § 2106.04(a)(2)(III).
Therefore, the independent claim recites a judicial exception.
Step 2A Prong Two: Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, the judicial exception recited above is not integrated into a practical application. The claims recite the following additional elements, but these additional elements are not sufficient to integrate the judicial exception into a practical application:
a support processing node being configured to apply influence balancing to a plurality of data inputs to produce combined influence data; (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or generally links exception to a technological environment)
and an artificial intelligence node being configured to receive as input the combined influence data that influences output generation. (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or generally links exception to a technological environment)
The “apply” step requires a “support processing node” to apply influence balancing to data inputs. The support processing node is used to apply the recited judicial exception without placing any limitation on how the support processing node operates. The limitation amounts to mere instructions to “apply” the judicial exception on a computer. It can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
The “and an artificial intelligence node being configured to receive as input …” step is recited at a high-level of generality such that the limitation amounts to no more than mere instructions to “apply” the judicial exception on a computer. It can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
Therefore, the above limitations do not integrate the judicial exception into a practical application.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception?
No. The claims do not include additional elements that are sufficient for the claims to amount to significantly more than the judicial exception.
In regards to the “support processing node” in the “apply” step, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
In regards to the “and an artificial intelligence node being configured to receive as input …” step, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
Therefore, independent claim 29 is not patent eligible.
Independent Claim 34
Step 2A Prong One: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, independent claim 34, under the broadest reasonable interpretation, recites the following limitations that are abstract ideas:
and an assisting node being configured to produce, based on the second output, influence data that is applied to influence the first neural network node. (mental process)
The “produce” step involves using second output to determine influence data which amounts to no more than observations, evaluations, and judgments that can be performed in the human mind or with the use of a physical aid (e.g., pen and paper). The claim recites the step of producing influence data at a high degree of generality, thus the step is not required to have any specific level of complexity that would preclude the step from being mental processes. Therefore, the “produce” step is considered to be mental processes, see MPEP § 2106.04(a)(2)(III).
Therefore, the independent claim recites a judicial exception.
Step 2A Prong Two: Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, the judicial exception recited above is not integrated into a practical application. The claims recite the following additional elements, but these additional elements are not sufficient to integrate the judicial exception into a practical application:
a first neural network node configured to produce a first output of a first media type; (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or generally links exception to a technological environment)
a second neural network node configured to produce a second output of a second media type, the first media type being different the second media type; (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or generally links exception to a technological environment)
and an assisting node being configured to produce, based on the second output, influence data that is applied to influence the first neural network node. (MPEP § 2106.05(f) mere instructions to implement an abstract idea on a computer, or generally links exception to a technological environment)
The “a first neural network node configured to …” step is recited at a high-level of generality such that the limitation amounts to no more than mere instructions to “apply” the judicial exception on a computer. It can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
The “a second neural network node configured to …” step is recited at a high-level of generality such that the limitation amounts to no more than mere instructions to “apply” the judicial exception on a computer. It can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
The “produce” step requires an “assisting node” to produce influence data. The assisting node is used to apply the recited judicial exception without placing any limitation on how the assisting node operates. The limitation amounts to mere instructions to “apply” the judicial exception on a computer. It can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
Therefore, the above limitations do not integrate the judicial exception into a practical application.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception?
No. The claims do not include additional elements that are sufficient for the claims to amount to significantly more than the judicial exception.
In regards to the “a first neural network node configured to …” step, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
In regards to the “a second neural network node configured to …” step, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
In regards to the “assisting node” in the “produce” step, the limitations are recited so generically such that they amount to no more than mere instructions to “apply” the judicial exception on a computer using generic computer components. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
Therefore, independent claim 34 is not patent eligible.
Dependent Claims 20-23, 25-28, 30-33 and 35-38
The remaining dependent claims being rejected do not recite additional elements, whether considered individually or in combination, that are sufficient to integrate the judicial exception into a practical application or amount to significantly more than a judicial exception.
Regarding Claims 20, 25, 30 and 35,
(claim 20) the plurality of influence data comprises personal data influences, and the processing circuitry is operable to merge the personal data influences with user queries.
(claim 25) the support node is operable to merge personal data influences with user queries.
(claim 30) the plurality of data inputs comprises personal data influences, and the artificial intelligence node is operable to merge the personal data influences with user queries.
(claim 35) the influence data comprises personal data influences, and the assisting node is operable to merge the personal data influences with user queries.
The “plurality of influence data comprises personal data influences” steps are recited at a high-level of generality such that the limitations amount to no more than mere instructions to “apply” the judicial exception on a computer. They can also be viewed as nothing more than attempts to generally link the use of the judicial exception to the technological environment of computers. Mere instructions to apply a judicial exception cannot provide an inventive concept. See MPEP § 2106.05(f).
The “merge” steps are mental processes akin to a human evaluation/judgment/observation. See MPEP § 2106.04(a)(2)(III).
The “merge” step of Claim 20 requires “processing circuitry” to merge personal influences. The processing circuitry is used to apply the recited judicial exception without placing any limitation on how the processing circuitry operates. The “merge” step of Claim 25 requires a “support node” to merge personal influences. The support node is used to apply the recited judicial exception without placing any limitation on how the support node operates. The “merge” step of Claim 30 requires an “artificial intelligence node” to merge personal influences. The artificial intelligence node is used to apply the recited judicial exception without placing any limitation on how the artificial intelligence node operates. The “merge” step of Claim 35 requires an “assisting node” to merge personal influences. The assisting node is used to apply the recited judicial exception without placing any limitation on how the assisting node operates.
Therefore, the limitations amount to mere instructions to “apply” the judicial exception on a computer. The limitations can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
Regarding Claims 21, 26, 31 and 36,
(Claim 21) the processing circuitry is operable to: select a fixed merger topology, and combine influences, from personal data and user input requests, through pre-weighted influence balancing before the merger operation.
(Claim 26) the support node is operable to: select a fixed merger topology, and combine influences, from personal data and user input requests, through pre-weighted influence balancing.
(Claim 31) the artificial intelligence node is operable to: select a fixed merger topology, and combine influence, from personal data and user input request, through pre-weighted influence balancing.
(Claim 36) the assisting node is operable to: select a fixed merger topology, and combine influence, from personal data and user input request, through pre-weighted influence balancing.
The “select” and “combine” steps are mental processes akin to a human evaluations / judgments /observations. See MPEP § 2106.04(a)(2)(III).
The “select” step of Claim 21 requires “processing circuitry” to select a fixed merger topology. The processing circuitry is used to apply the recited judicial exception without placing any limitation on how the processing circuitry operates. The “select” step of Claim 26 requires a “support node” to select a fixed merger topology. The support node is used to apply the recited judicial exception without placing any limitation on how the support node operates. The “select” step of Claim 31 requires an “artificial intelligence node” to select a fixed merger topology. The artificial intelligence node is used to apply the recited judicial exception without placing any limitation on how the artificial intelligence node operates. The “select” step of Claim 36 requires an “assisting node” to select a fixed merger topology. The assisting node is used to apply the recited judicial exception without placing any limitation on how the assisting node operates.
Therefore, the limitations amount to mere instructions to “apply” the judicial exception on a computer. The limitations can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
Regarding Claims 22, 27, 32 and 37,
(Claim 22) the processing circuitry is operable to: process influence inputs from personal data, remote sources, and current events, and adjust an impact, of the personal data, the remote sources, and the current events, according to user preferences.
(Claim 27) the support node is operable to: process influence inputs from personal data, remote sources, and current events, and adjust an impact, of the personal data, the remote sources, and the current events, according to user preferences.
(Claim 32) the artificial intelligence node is operable to: process influence inputs from personal data, remote sources, and current events, and adjust an impact, of the personal data, the remote sources, and the current events, according to user preferences.
(Claim 37) the assisting node is operable to: process influence inputs from personal data, remote sources, and current events, and adjust an impact, of the personal data, the remote sources, and the current events, according to user preferences.
The “process influence inputs” steps are mental processes akin to human evaluations / judgments /observations. The “adjust” steps are mental processes akin to human evaluations / judgments / observations. See MPEP § 2106.04(a)(2)(III).
The “process influence inputs” step of Claim 22 requires “processing circuitry” to process influence inputs. The processing circuitry is used to apply the recited judicial exception without placing any limitation on how the processing circuitry operates. The “process influence inputs” step of Claim 27 requires a “support node” to process influence inputs. The support node is used to apply the recited judicial exception without placing any limitation on how the support node operates. The “process influence inputs” step of Claim 32 requires an “artificial intelligence node” to process influence inputs. The artificial intelligence node is used to apply the recited judicial exception without placing any limitation on how the artificial intelligence node operates. The “process influence inputs” step of Claim 37 requires an “assisting node” to process influence inputs. The assisting node is used to apply the recited judicial exception without placing any limitation on how the assisting node operates.
Therefore, the limitations amount to mere instructions to “apply” the judicial exception on a computer. The limitations can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
Regarding Claims 23, 28, 33 and 38,
(Claim 23) the processing circuitry is operable to: evaluate and adjust the plurality of influence data to maintain context and correlation in an output according to a baseline.
(Claim 28) the support node is operable to: evaluate and adjust a plurality of influence data to maintain context and correlation in an output according to a baseline.
(Claim 33) the artificial intelligence node is operable to: evaluate and adjust the plurality of data inputs to maintain context and correlation in an output according to a baseline.
(Claim 38) the assisting node is operable to: evaluate and adjust the plurality of influence data to maintain context and correlation in the first output and the second output according to a baseline.
The “evaluate and adjust” steps are mental processes akin to human evaluations / judgments / observations. See MPEP § 2106.04(a)(2)(III).
The “evaluate and adjust” step of Claim 23 requires “processing circuitry” to evaluate and adjust influence data. The processing circuitry is used to apply the recited judicial exception without placing any limitation on how the processing circuitry operates. The “evaluate and adjust” step of Claim 28 requires a “support node” to evaluate and adjust influence data. The support node is used to apply the recited judicial exception without placing any limitation on how the support node operates. The “evaluate and adjust” step of Claim 33 requires an “artificial intelligence node” to evaluate and adjust influence data. The artificial intelligence node is used to apply the recited judicial exception without placing any limitation on how the artificial intelligence node operates. The “evaluate and adjust” step of Claim 38 requires an “assisting node” to evaluate and adjust influence data. The assisting node is used to apply the recited judicial exception without placing any limitation on how the assisting node operates.
Therefore, the limitations amount to mere instructions to “apply” the judicial exception on a computer. The limitations can also be viewed as nothing more than an attempt to generally link the use of the judicial exception to the technological environment of computers, see MPEP § 2106.05(f).
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
The following is the reference relied upon in the rejections below:
Cao, Qian, et al. "Multi-modal experience inspired AI creation." Proceedings of the 30th ACM International Conference on Multimedia. 2022.
Claims 19, 23-24, 28-29, 33-34 and 38 are rejected under 35 U.S.C. 102a(1) as being anticipated by Cao.
Regarding Claim 19, Cao teaches:
An electronic infrastructure, comprising ((P. 1445, Abstract) “we firstly design a multi-channel sequence-to-sequence architecture equipped with a multi-modal attention network.”):
a plurality of influence data originating from a corresponding plurality of data sources ((P. 1446, Sec. 1, ¶2) “we design a multi-channel sequence to-sequence architecture, where different modalities are firstly projected into the same space based on the attention mechanism, and then the output is recurrently generated based on the fused information. In order to control the influence of the input on the output, we design a novel self-attention method to enable the incorporation of heuristic prior knowledge.”
(P. 1446, Sec. 3, ¶1) “the raw images and texts are firstly handled by a multi-channel sequence processor to produce their semantic embeddings.”);
and processing circuitry operable to perform influence balancing and merger operations on the plurality of influence data to produce combined influence data destined to influence an operational neural network (The Examiner interprets “influence balancing” according to its broadest reasonable interpretation (BRI) in view of the Applicant’s specification as encompassing an attention mechanism. This interpretation is consistent with the illustrative descriptions in the Applicant’s specification, (see excerpts below).
Applicant’s written description at (P. 32, Lines 5-10): “influence balancing includes decreasing (even to the point of selectively blocking) and increasing influence contributions of a plurality of influence data from a plurality of influence sources. For example, with five influence inputs, an influence balancing topology node may fully block the first influence input, leave the second and third influence inputs untouched, decrease the intensity of the fourth influence input, and elevate the intensity of the fifth influence input.”
(P. 1447, Sec. 3.2, ¶2) “if the distance between the input and output is large, then the influence should be small. For example, although image C in Figure 1 influences the third and fourth sentences, the first sentence is not that relevant to it. In order to encode such intuitions into our model, we further introduce a regularizer to constrain the attention weights. … we regularize the attention weights with a prior, which encodes the intuition that a larger input-output distance should lead to a lower influence between them.”
(P. 1446, Sec. 3, ¶1) “the raw images and texts are firstly handled by a multi-channel sequence processor to produce their semantic embeddings. Then, the embedding at each step is separated into different parts to influence the final output. At last, different modalities are fused with an attention network. The last part is the decoder, aiming to predict the final output sentences.”
See Figure 2 on P. 1447 depicting a Multi-Modal Fusion Network is used to fuse embeddings of topic, image, and text.
Raw images and texts (‘plurality of influence data’) have their influence on an output adjusted by using attention weights (attention mechanism). The influence of each image and text is adjusted to control the influence of each image or text on a recurrently generated output (see P. 1446, Sec. 1, ¶2). The inputs are then fused to generate fused information (see P. 1446, Sec. 1, ¶2). Therefore, adjusting influences of input images and texts by an attention mechanism is performing influence balancing, and fusing the input images and texts into fused information is performing merger (fusion) operations on influence data to produce combined influence data (fused information). A decoder (‘operational neural network’) then uses the fused information to generate output sentences, therefore the fused information influences the decoder’s outputs. Processing circuitry is implied by using a Multi-Modal Fusion Network to fuse embeddings.).
Regarding Claims 23 and 28, Cao teaches:
The electronic infrastructure of claim 24, wherein: the support node is operable to: evaluate and adjust a plurality of influence data to maintain context and correlation in an output according to a baseline ((P. 1447, Sec. 3.2, ¶2) “if the distance between the input and output is large, then the influence should be small … we further introduce a regularizer to constrain the attention weights. Formally, we minimize the distance between
α
j
… and a pre-defined distribution … where we implement
γ
(
j
)
as the Gaussian distribution, … By minimizing
α
j
and
γ
(
j
)
, we regularize the attention weights with a prior, which encodes the intuition that a larger input-output distance should lead to a lower influence between them.”
(P. 1446, Sec. 3, ¶1) “the raw images and texts are firstly handled by a multi-channel sequence processor to produce their semantic embeddings. Then, the embedding at each step is separated into different parts to influence the final output. At last, different modalities are fused with an attention network. The last part is the decoder, aiming to predict the final output sentences.”
Input raw images and texts are transformed into semantic embeddings (‘influence data’). Attention weights are compared to a pre-defined Gaussian distribution (‘baseline’) to minimize (regularize) the attention weights so that the resulting attention weight values reflect that input semantic embeddings with larger distances to an output have less influence on the output. Influence of an input semantic embedding is therefore evaluated and adjusted by attention weights regularized according to a baseline. By adjusting the semantic embedding’s influence, the input’s influence (contribution) to the output is adjusted, therefore adjusting semantic embeddings (‘influence data’). The self-attention module (‘supporting node’) controls the influence of input semantic embeddings on an output (see P. 1446, Sec. 1, ¶2), therefore the embeddings with smaller distances to an output represent embeddings that are semantically correlated and maintain context to an output.).
Regarding Claim 24, Cao teaches:
An artificial intelligence infrastructure, comprising ((P. 1445, Abstract) “we firstly design a multi-channel sequence-to-sequence architecture equipped with a multi-modal attention network.”):
a first neural network node ((P. 1446, Sec. 3, ¶1) “The last part is the decoder, aiming to predict the final output sentences.”
See Figure 2 on P. 1447 depicting an Experience Enhanced Sentence Decoder is comprised of a GPT-2 model (‘first neural network node’).);
a second neural network node ((P. 1448, Sec. 3.6, ¶2) “To capture the sequential information of the experiences, we process the embedded image and text with the Gated Recurrent Units (GRU) [8]. The input image sequence … are encoded”
See Figure 2 on P. 1447 depicting an Image GRU used to encode input images.);
and a support node being configured to respond to first data generated by the first neural network node by producing, from at least the first data, second data that is used to influence the first neural network node (The Examiner interprets “support node” according to its BRI in view of the Applicant’s specification as encompassing a self-attention module. This interpretation is consistent with the illustrative descriptions in the Applicant’s specification, (see excerpts below).
Applicant’s written description at (P. 22, Lines 9-12): “Each instance of each support processing element S1-S17 221-273 are designed to service a particular needed function at least once in at least one node … Such support processing elements are referred to herein as a "support processing topology node" or merely a "support processing node."”
Applicant’s written description at (P. 24, Lines 18-20): “S7 241 might then respond the S4 231 identified relationships and create a merged influence output that balances and weighs the value of one of such text inputs against the value of others.”
Applicant’s written description at (P. 20, Lines 16-19): “wherein such support processing code provides assistance to one or many AI elements … Support processing elements such as S1 211 are stored within the support processing code 167 (Fig. 1), and they execute on the software processing units 111”
(P. 1447, Sec. 3.2, ¶1-2) “the image or text may influence a span of the output sequence. … we design a tailored module to capture the spanning influence of the input on the outputs. Specifically, we let the hidden embedding derived in the above section influence the output sequences attentively. … if the distance between the input and output is large, then the influence should be small”
(P. 1446, Sec. 1, ¶2) “we design a multi-channel sequence to-sequence architecture, where different modalities are firstly projected into the same space based on the attention mechanism, and then the output is recurrently generated based on the fused information. In order to control the influence of the input on the output, we design a novel self-attention method to enable the incorporation of heuristic prior knowledge.”
An output sequence (sentence) is generated by a decoder (‘first neural network node’), see (P. 1446, Sec. 3, ¶1), therefore an output sequence is ‘first data’. A self-attention module fuses different modalities represented by embeddings to generate fused information (‘second data’). The fused information represents the combined influence each input has on an output sequence (‘first data’), therefore fused information is merged influence output that balances and weighs the values of each input, and the self-attention module responds to ‘first data’ by producing from the first data (output sequence), second data. The fused information is then used by the decoder to recurrently generate output sequences, therefore using second data (fused information) to influence the first neural network node’s (decoder) outputs.
The self-attention module assists the decoder (an AI element) by generating fused information for the decoder, and the fused information is generated by fusing influences of different inputs. Therefore, the self-attention module is a support processing element (module) that generates merged influence output to assist a decoder in generating outputs, and therefore the module is a ‘support node’.).
Regarding Claim 29, the rejection of claim 24 is incorporated. The difference in scope being:
a support processing node being configured to apply influence balancing to a plurality of data inputs to produce combined influence data ((P. 1447, Sec. 3.2, ¶1-2) “the image or text may influence a span of the output sequence. … we design a tailored module to capture the spanning influence of the input on the outputs. Specifically, we let the hidden embedding derived in the above section influence the output sequences attentively. … if the distance between the input and output is large, then the influence should be small”
(P. 1446, Sec. 1, ¶2) “we design a multi-channel sequence to-sequence architecture, where different modalities are firstly projected into the same space based on the attention mechanism, and then the output is recurrently generated based on the fused information. In order to control the influence of the input on the output, we design a novel self-attention method to enable the incorporation of heuristic prior knowledge.”
A self-attention module (‘support processing node’) fuses different modalities (‘plurality of data inputs’) represented by embeddings to generate fused information (‘combined influence data’). The influence of each input image and text is adjusted to control the influence each input has on an output sequence. Therefore, adjusting influences of input images and texts by a self-attention module is performing influence balancing, and the fused information represents the combined, weighted influence each input has on an output sequence.);
and an artificial intelligence node being configured to receive as input the combined influence data that influences output generation ((P. 1446, Sec. 3, ¶1) “The last part is the decoder, aiming to predict the final output sentences.”
(P. 1446, Sec. 1, ¶2) “the output is recurrently generated based on the fused information. In order to control the influence of the input on the output, we design a novel self-attention method”
Fused information (‘combined influence data’) is used by a decoder (‘artificial intelligence node’) to recurrently generate output sequences (sentences), therefore using combined influence data to influence output generation of the decoder.).
Regarding Claim 33, Cao teaches:
The electronic infrastructure of claim 29, wherein: the artificial intelligence node is operable to: evaluate and adjust the plurality of data inputs to maintain context and correlation in an output according to a baseline (((P. 1447, Sec. 3.2, ¶2) “if the distance between the input and output is large, then the influence should be small … we further introduce a regularizer to constrain the attention weights. Formally, we minimize the distance between
α
j
… and a pre-defined distribution … where we implement
γ
(
j
)
as the Gaussian distribution, … By minimizing
α
j
and
γ
(
j
)
, we regularize the attention weights with a prior, which encodes the intuition that a larger input-output distance should lead to a lower influence between them.”
(P. 1446, Sec. 3, ¶1) “the raw images and texts are firstly handled by a multi-channel sequence processor to produce their semantic embeddings. Then, the embedding at each step is separated into different parts to influence the final output. At last, different modalities are fused with an attention network. The last part is the decoder, aiming to predict the final output sentences.”
(P. 1446, Sec. 1, Last Paragraph) “we design a neural sequence-to sequence model and also enhance it with a prior knowledge guided self-attention module”
Input raw images and texts are transformed into semantic embeddings (‘plurality of data inputs’). Attention weights are compared to a pre-defined Gaussian distribution (‘baseline’) to minimize (regularize) the attention weights so that the resulting attention weight values reflect that input semantic embeddings with larger distances to an output have less influence on the output. Influence of an input semantic embedding is therefore evaluated and adjusted by attention weights regularized according to a baseline. By adjusting the semantic embedding’s influence, the input’s influence (contribution) to the output is adjusted, therefore adjusting semantic embeddings (‘plurality of data inputs’). The self-attention module controls the influence of input semantic embeddings on an output (see P. 1446, Sec. 1, ¶2), therefore the embeddings with smaller distances to an output represent embeddings that are semantically correlated and maintain context to an output.
A neural sequence-to-sequence model (‘artificial intelligence node’) is enhanced with a self-attention module, therefore the model is operable to adjust influences of inputs.).
Regarding Claim 34, the rejection of claim 24 is incorporated. The difference in scope being:
a first neural network node configured to produce a first output of a first media type ((P. 1446, Sec. 3, ¶1) “The last part is the decoder, aiming to predict the final output sentences.”
See Figure 2 on P. 1447 depicting an Experience Enhanced Sentence Decoder is comprised of a GPT-2 model (‘first neural network node’).);
a second neural network node configured to produce a second output of a second media type ((P. 1448, Sec. 3.6, ¶2) “To capture the sequential information of the experiences, we process the embedded image and text with the Gated Recurrent Units (GRU) [8]. The input image sequence … are encoded … we deploy normalization upon the outputs. … The normalized embeddings are then fed into spanning influence modeling layers.”
See Figure 2 on P. 1447 depicting an Image GRU used to encode input images.),
the first media type being different the second media type (A decoder generates sentences (text) and an Image GRU generates image encodings, therefore the decoder and Image GRU generate outputs of different media types.);
and an assisting node being configured to produce, based on the second output, influence data that is applied to influence the first neural network node ((P. 1447, Sec. 3.2, ¶1-2) “the image or text may influence a span of the output sequence. … we design a tailored module to capture the spanning influence of the input on the outputs. Specifically, we let the hidden embedding derived in the above section influence the output sequences attentively. … if the distance between the input and output is large, then the influence should be small”
(P. 1446, Sec. 1, ¶2) “we design a multi-channel sequence to-sequence architecture, where different modalities are firstly projected into the same space based on the attention mechanism, and then the output is recurrently generated based on the fused information. In order to control the influence of the input on the output, we design a novel self-attention method to enable the incorporation of heuristic prior knowledge.”
An Image GRU processes embedded images to generate normalized embeddings that are used when calculating input influence (see P. 1448, Sec. 3.6, ¶2), therefore the embeddings are ‘second output’. A self-attention module (‘assisting node’) fuses different modalities represented by normalized text and image embeddings (‘second output’) to generate fused information. The influence of each input image and text is adjusted by the module to control the influence each input has on an output sequence, and then the influences along with their inputs are fused into fused information (‘influence data’). The fused information is used by a decoder (‘first neural network node’) to recurrently generate output sequences (sentences), therefore applying ‘influence data’ to influence output generation of the decoder.).
Regarding Claim 38, Cao teaches:
The electronic infrastructure of claim 34, wherein: the assisting node is operable to: evaluate and adjust the plurality of influence data to maintain context and correlation in the first output and the second output according to a baseline ((P. 1447, Sec. 3.2, ¶2) “if the distance between the input and output is large, then the influence should be small … we further introduce a regularizer to constrain the attention weights. Formally, we minimize the distance between
α
j
… and a pre-defined distribution … where we implement
γ
(
j
)
as the Gaussian distribution, … By minimizing
α
j
and
γ
(
j
)
, we regularize the attention weights with a prior, which encodes the intuition that a larger input-output distance should lead to a lower influence between them.”
(P. 1446, Sec. 3, ¶1) “the raw images and texts are firstly handled by a multi-channel sequence processor to produce their semantic embeddings. Then, the embedding at each step is separated into different parts to influence the final output. At last, different modalities are fused with an attention network. The last part is the decoder, aiming to predict the final output sentences.”
See (P. 1447, Sec. 3.2, ¶1-2) describing distances between output sequences and inputs are used to determine influence.
The self-attention module (‘assisting node’) controls the influence of input semantic embeddings on an output sequence (see P. 1446, Sec. 1, ¶2). Input texts and raw images are transformed into semantic embeddings. Attention weights are compared to a pre-defined Gaussian distribution (‘baseline’) to minimize (regularize) the attention weights so that the resulting attention weight values reflect that input semantic embeddings with larger distances to an output sequence (‘first output’) have less influence on the output sequence. Influence of an input semantic embedding is therefore evaluated and adjusted by attention weights regularized according to a baseline.
The input text semantic embeddings and image semantic embeddings (‘second output’) weighted by influences are fused into fused information (see P. 1446, Sec. 1, ¶2). Therefore, when each input’s influence is adjusted, fused information (‘plurality of influence data’) which is weighted by influences is also adjusted. The fused information contains the influence of each input, therefore, the fused information represents how much each semantic embedding influences an output sequence (‘first output’). An image semantic embedding (‘second output’) with a smaller distance to an output sequence (‘first output’) will have a large influence on the output sequence, therefore the large influence represents a semantic embedding (‘second output’) that is semantically correlated and maintains context to the output sequence.).
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 following are the references relied upon in the rejections below:
Wong, Yongkang, et al. "Compute to tell the tale: Goal-driven narrative generation." Proceedings of the 30th ACM International Conference on Multimedia. 2022.
Claims 20, 22, 25, 27, 30, 32, 35 and 37 are rejected under 35 U.S.C. 103 as being unpatentable over Cao / Wong.
Regarding Claims 20, 25, 30 and 35, Cao teaches: The electronic infrastructure of claim 29, however Cao does not teach merging personal data influences with user queries, which is taught by Wong:
wherein: the plurality of data inputs comprises personal data influences, and the artificial intelligence node is operable to merge the personal data influences with user queries ((P. 6875, Abstract) “Informed by the real-world impact of narrative generation, we then illustrate several practical use cases in Video Logging as a Service platform which enables users to get more out of the data through a goal-driven intelligent storytelling AI agent.”
(P. 6877, Sec. 3.2.1) “Query Encoder. Queries are important inputs to the framework reflecting the users’ intention on which part of information is of interest and the form the narrative is expected to take. They are innately unstructured and multimodal, which calls for effective encoding and fusion methods.”
(P. 6878, Sec. 3.3.2) “Story templates are schemes that control how narrative elements are arranged in the generated narrative, … The story templates are expected to not only guarantee logical correctness, but also ensure a aesthetically enjoyable narrative based on user profile. This requires learning different literature styles from a large text corpus, as well as take into account user query and other external knowledge.”
(P. 6879, Sec. 5.2.3) “it is necessary to consider the user profile and preference, circumstances, and target audience during the narrative generation process. On one hand, such personalization can be based on user’s preference on genre, style, or aesthetics. … These factors have to be jointly considered during the narrative structure modeling and narrative generation.”
A goal-driven intelligent storytelling AI agent (‘artificial intelligence node’) encodes input queries and fuses the queries with user preferences (‘personal data influences’) to generate a narrative that takes into account a user’s intention (derived from a query) and preferences. The agent jointly considers user preference and other knowledge to generate a narrative that aligns with a user’s intention (query), therefore merging user preferences with user queries to generate AI responses.).
Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to modify the method of Cao with the technique disclosed by Wong to merge personal data and user queries when generating responses. By merging personal data and user queries when generating responses, a user’s personal data and intentions can be combined to generate a response that is both personalized and relevant, thereby enhancing user engagement and satisfaction.
Regarding Claims 22, 27, 32 and 37, Cao teaches The electronic infrastructure of claim 29, however Cao does not teach adjusting an impact of personal data, remote sources, and current events according to user preferences, which is taught by Wong:
wherein: the artificial intelligence node is operable to: process influence inputs from personal data, remote sources, and current events ((P. 6875, Abstract) “Informed by the real-world impact of narrative generation, we then illustrate several practical use cases in Video Logging as a Service platform which enables users to get more out of the data through a goal-driven intelligent storytelling AI agent.”
(P. 6878, Sec. 3.3.1) “User Profile. The users’ demographic information and interests are important sources from which the users’ preference on the output narrative can be inferred. In the context of this work, the profile affects the user’s expectation on the narrative: the factual contents it delivers, the arguments it proposes, the literature style it uses, etc.”
(P. 6878, Sec. 3.3.3) “Domain knowledge is therefore an indispensable part of a successful NG framework. Existing knowledge databases provide domain knowledge using highly structured graphs or logic constraints. For example, Wikidata [77] provides common sense knowledge, while ConceptNet [74] provides taxonomy information between words.”
(P. 6879, Sec. 5.2.3) “user’s privacy concern would determine what elements can be included. For example, if the narrative is generated for family members, more private information can be included as opposed to news article.”
A goal-driven intelligent storytelling AI agent processes a user’s profile (‘personal data’), domain knowledge from existing knowledge databases (‘remote sources’), and news articles (‘current events’) as inputs to generate a narrative. User profile, domain knowledge, and news articles affect how narratives are generated, therefore the inputs are influence inputs.),
and adjust an impact, of the personal data, the remote sources, and the current events, according to user preferences ((P. 6879, Sec. 5.2.3) “Although the factual narrative is based on the actual event with underlying temporal and causal relation, a good narrative is not bounded by a single objective story and enjoys various forms depending on the user’s preference. Therefore, it is necessary to consider the user profile and preference, circumstances, and target audience during the narrative generation process. On one hand, such personalization can be based on user’s preference on genre, style, or aesthetics. On the other hand, user’s privacy concern would determine what elements can be included. For example, if the narrative is generated for family members, more private information can be included as opposed to news article. In addition, narrative persuasion would evoke various type of emotions. These factors have to be jointly considered during the narrative structure modeling and narrative generation.”
User preferences (genre, style, aesthetics, privacy concerns) affect how narratives are generated. When a user has low privacy concerns, a narrative is generated that uses more demographic / private information from a user profile (family member information), therefore the impact user profile (‘personal data’) has on narrative generation is adjusted according to user preferences. When a user has low privacy concerns, less information from a news article is included in a generated narrative, therefore the impact a news article (‘current events’) has on narrative generation is adjusted according to user preferences. User preferences affect which elements can be included in a generated narrative, therefore the impact of domain knowledge (‘remote sources’) is also adjusted according to user preferences.).
Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to modify the method of Cao with the technique disclosed by Wong to adjust the impact personal data, remote sources, and current events have on a generated response according to user preferences. By adjusting the impact personal data, remote sources, and current events have on a generated response according to user preferences, a user’s preferences can be used to determine how personal, relevant, and current a generated response will be, thereby generating highly customized responses that are likely to fulfill a user’s needs.
The following are the references relied upon in the rejections below:
Yang, Hao, et al. "Exploring the potential of large language models in personalized diabetes treatment strategies." medRxiv (2023): 2023-06.
Claims 21, 26, 31 and 36 are rejected under 35 U.S.C. 103 as being unpatentable over Cao / Yang.
Regarding Claims 21, 26, 31, and 36, Cao teaches: The electronic infrastructure of claim 19, however Cao does not teach selecting a fixed merger topology and pre-weighted influence balancing, which is taught by Yang:
wherein: the processing circuitry is operable to: select a fixed merger topology (The Examiner interprets “select a fixed merger topology” according to its BRI in view of the Applicant’s specification as encompassing selecting a base model for specific users. This interpretation is consistent with the illustrative descriptions in the Applicant’s specification, (see excerpts below).
Applicant’s written description at (P. 33, Lines 17-20) “a first fixed merger topology node selected from a plurality of fixed merger support processing topology nodes with each of such nodes being designed to service a different user class, and wherein the first fixed merger topology node being configured to service only the user class in which the current user is considered a member.”
Applicant’s written description at (P. 35, Lines 15-22) “topology nodes can be dedicated to service a particular user class or can adapt to service all user classes. A user through their own historical interactions can define their own personal user class with only one member, themselves. … One user may have a set of personalized topology specifications with a certain number and layout of nodes that are unique to that user or at least unique to a class of users.”
(P. 5, Sec. Fine-tune, ¶1) “We will use ChatGLM as the base model and fine-tune it for specific tasks in the field of diabetes management. For the fine-tuning process, we will employ commonly used techniques for LLMs, namely P-tuning [12] and LoRA [13]. These methods have been proven effective in adapting pre-trained language models to domain-specific tasks and improving their performance”
(P. 8, Sec. Model Integreated) “The fine-tuned model will be integrated into the EHR system (Figure 4) to facilitate the management of diabetes patients by clinical physicians. This integration will enable the provision of personalized treatment recommendations for diabetes patients.”
A ChatGLM model is selected as a base model and the model is fine-tuned for diabetic patients. ChatGLM is an AI model, therefore it is a node and therefore a topology node. A fine-tuned ChatGLM is fine-tuned to only generate treatment recommendations for diabetic patients (user class), therefore the fine-tuned model is ‘fixed’ and therefore is a fixed topology node. A fine-tuned ChatGLM merges a new domain (diabetic patients) with its pre-trained domain, therefore the model is a ‘fixed merger topology’ that is adapted to service diabetic patients (user class) by generating personalized treatment recommendations. Therefore, selecting ChatGLM for adaptation to diabetic patients is ‘selecting a fixed merger topology’.),
and combine influences, from personal data and user input requests, through pre-weighted influence balancing before the merger operation ((P. 6, Sec. LoRA) “Lora (low-rank adaptation) is a low-rank adaptation technique that allows for a significant reduction in the number of trainable parameters in downstream tasks while keeping the weights of the pre-trained model unchanged”
(P. 11, Last Paragraph) “when a physician completes the patient's basic information, chief complaints, history of present illness and diagnosis, the system automatically collects the information and uses a fine-tuned GLM to generate appropriate treatment recommendations.”
(P. 3-4, Sec. Methods) “The collected data will include information such as the patient's department of visit, age, gender, chief complaint, present illness history, and diagnosis, which will serve as input for the model.”
Pre-trained ChatGLM is fine-tuned using low-rank adaptation (LoRA). LoRA keeps some weights of the ChatGLM model unchanged, therefore the unchanged weights are pre-weighted. When LoRA fine-tunes ChatGLM, LoRA uses collected data (see P. 3, Sec. Introduction, Last Paragraph). Collected data includes a patient’s basic information (‘personal data’) and chief complaint (‘user input requests’). Patient’s basic information and chief complaint are used to generate a treatment recommendation, therefore patient’s basic information and chief complaint are influences. When the model is being fine-tuned, unchanged weights and fine-tuned weights are used to weigh source domain information (the domain ChatGLM was pre-trained on) and new domain information (patient’s basic information and chief complaint), therefore balancing source domain information and new domain information by using pre-weighted weights (and therefore combining influences (source and new domain) through pre-weighted influence balancing before a merger operation (generating a treatment recommendation)).).
Before the effective filing date of the claimed invention, it would have been obvious to a person of ordinary skill in the art to modify the method of Cao with the technique disclosed by Yang to select and adapt a model to serve a specific user class. By selecting and adapting a model to serve a specific user class, a model can be tuned to tailor its responses to users of a certain class, thereby providing personalized responses and improving relevance of generated outputs.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Wang et al. (“Generative recommendation: Towards next-generation recommender paradigm”) teaches fusing user instructions, user feedback, an existing item, and knowledge from the web to refine or generate custom content.
Kim et al. (“A multi-modal story generation framework with AI-driven storyline guidance”) teaches generating a multi-modal story by having one machine learning model generate a paragraph and a second machine learning model generate a storyline. Each model’s outputs are used by the other model to generate the next component of the story in a cohesive manner. Once the entire story is generated, an image visualization model generates images to accompany each scene of the story.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PEDRO J MORALES whose telephone number is (571)272-6106. The examiner can normally be reached 8:30 AM - 6:00 PM.
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, MIRANDA M HUANG can be reached at (571)270-7092. 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.
/PEDRO J MORALES/Examiner, Art Unit 2124
/MIRANDA M HUANG/Supervisory Patent Examiner, Art Unit 2124