DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
This action is responsive to remarks filed 11/21/2025. Claims 1–13 are amended. No claims are cancelled, and claim 16 is new.
Claims 1–16 are pending for examination.
Response to Arguments
In reference to 35 USC § 112
Applicant’s arguments and amendments, filed on 11/21/2025, with respect to the §112(b) rejections have been fully considered and are persuasive. Thus, the §112(b) rejections are withdrawn.
In reference to 35 USC § 101
Applicant’s arguments, filed on 11/21/2025, with respect to the § 101 rejections have been fully considered but are not persuasive. Applicant argues, beginning on Pg. 9 in the Remarks, that “the Office has not met its burden of establishing that the claims are directed to an abstract idea or that the claims do not recite significantly more.” Specifically, Applicant argues that the amended claims “are not directed to an abstract idea, have a practical application, and contain a number of particular limitations which ensure that no underlying idea would be tied up.” Examiner respectfully disagrees.
Applicant argues, beginning on Pg. 10 in the Remarks, that More specifically, Applicant argues that “if a human were to utilize pen and paper, they would never seek to utilize the library of expert linkers as they would perform the linking of the documents based on the raw documents themselves.” Examiner respectfully disagrees. See also MPEP 2106.04(a)(2)(III)(C):
Claims can recite a mental process even if they are claimed as being performed on a computer. The Supreme Court recognized this in Benson, determining that a mathematical algorithm for converting binary coded decimal to pure binary within a computer’s shift register was an abstract idea. The Court concluded that the algorithm could be performed purely mentally even though the claimed procedures "can be carried out in existing computers long in use, no new machinery being necessary." 409 U.S at 67, 175 USPQ at 675. See also Mortgage Grader, 811 F.3d at 1324, 117 USPQ2d at 1699 (concluding that concept of "anonymous loan shopping" recited in a computer system claim is an abstract idea because it could be "performed by humans without a computer").
In the instant case, as Applicant pointed out, if a human were to perform the method they would indeed be training their own mind (i.e., a neural network) to adapt to the library of expert linkers, which is entirely the same concept as Applicant’s claimed invention. Clearly, a person can assess documents, assign a score to them, and remember the scores they have assigned. Even if they could not clearly remember, they are afforded the pen and paper, which would allow them to make the connections. Examiner notes that the BRI of a vector could be as simple as a number with a plus (+) or minus (-) sign, representing direction along a single axis. Without more detail, Examiner maintains that the instant application includes abstract ideas including mental concepts that, with the aid of pen and paper, a human mind could perform.
Applicant argues, beginning on Pg. 11 in the Remarks, that with regard to the USPTO’s Example 39 Guidance, “the analysis specifically indicates that the claim, including this feature, are not directed to any judicial exceptions including a mental process because ‘the steps are not practically performed in the human mind.’” Examiner notes that Example 39 is directed to training a neural network for facial detection while the instant application is directed to generating accurate relationships among heterogenous documents in a semantic graph for an application, and the two are not analogous. Regardless, Example 39 is no directed to an abstract idea because there are abstract ideas asserted in the claims. Per the Office’s guidance, Example 39 includes collecting (insignificant extra-solution activity), applying (mere instructions to apply if there were an abstract idea), creating a training set (insignificant extra-solution activity), training the neural network (mere instructions to apply if there were an abstract idea), creating a second training set (insignificant extra-solution activity), and training the neural network a second time (mere instructions to apply if there were an abstract idea). Thus there are no abstract ideas. However, in contrast, the instant application includes generating a representation, computing a link score, and selecting targets based on the scores which are all found to be abstract ideas including mental concepts capable of being performed in the human mind.
Applicant argues, beginning of Pg. 12 in the Remarks, that “the claims provide a practical application under Prong Two of Step 2A of the Subject Matter Eligibility Test.” Examiner respectfully disagrees. See MPEP 2106.04(d)(1) and MPEP 2106.05(a) which respectively state:
if the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), the examiner should not determine the claim improves technology. Second, if the specification sets forth an improvement in technology, the claim must be evaluated to ensure that the claim itself reflects the disclosed improvement. That is, the claim includes the components or steps of the invention that provide the improvement described in the specification. The claim itself does not need to explicitly recite the improvement described in the specification (e.g., "thereby increasing the bandwidth of the channel").
An indication that the claimed invention provides an improvement can include a discussion in the specification that identifies a technical problem and explains the details of an unconventional technical solution expressed in the claim, or identifies technical improvements realized by the claim over the prior art. For example, in McRO, the court relied on the specification’s explanation of how the particular rules recited in the claim enabled the automation of specific animation tasks that previously could only be performed subjectively by humans, when determining that the claims were directed to improvements in computer animation instead of an abstract idea. McRO, 837 F.3d at 1313-14, 120 USPQ2d at 1100-01. In contrast, the court in Affinity Labs of Tex. v. DirecTV, LLC relied on the specification’s failure to provide details regarding the manner in which the invention accomplished the alleged improvement when holding the claimed methods of delivering broadcast content to cellphones ineligible. 838 F.3d 1253, 1263-64, 120 USPQ2d 1201, 1207-08 (Fed. Cir. 2016).
In the instant application, examiner contends the purported improvements are not improvements to a computer but are instead improvements directed to the abstract ideas themselves (i.e., the generating representations, computing link scores, and selecting targets) and therefore, do not integrate the abstract ideas into a practical application. Without details related to how the computer functions/technology have been improved, the abstract ideas are not integrated into a practical application and the additional elements to do not amount to significantly more. See detailed analysis of the newly amended claims in § 101 below.
Applicant argues, beginning on Pg. 13 in the Remarks, that “even if the Office believes that the claims are not patent-eligible under Step 2A and believes that the claims recite an abstract idea, which is not conceded, the claims should at least be found to be patent-eligible under Step 2B as the claims recite significantly more than a mental process.” Examiner notes MPEP 2106.07(a)II, which states in part:
if claim limitations that recite a generic computer component performing generic computer functions at a high level of generality, such as using the Internet to gather data, examiners can explain why these generic computing functions do not meaningfully limit the claim. Examiners should keep in mind that the courts have held computer-implemented processes to be significantly more than an abstract idea (and thus eligible), where generic computer components are able in combination to perform functions that are not merely generic. DDR Holdings, LLC v. Hotels.com, LP, 773 F.3d 1245, 1258-59, 113 USPQ2d 1097, 1106-07 (Fed. Cir. 2014). See MPEP § 2106.05(f) for more information about generic computing functions that the courts have found to be mere instructions to implement a judicial exception on a computer, and MPEP § 2106.05(d) for more information about well understood, routine, conventional activities and elements (a relevant consideration only in Step 2B).
In the instant application, examiner contends the purported improvements are not improvements to a computer but are instead improvements directed to the abstract ideas themselves (i.e., the generating representations, computing link scores, and selecting targets) and therefore, do not integrate the abstract ideas into a practical application and do not provide significantly more because they do not meaningfully limit the claims. Without details related to how the computer functions/technology have been improved and without providing meaningful limitation, the additional elements to do not amount to significantly more. See detailed analysis of the newly amended claims in § 101 below.
Thus, the § 101 rejections are maintained.
In reference to 35 USC § 103
Applicant’s arguments filed on 11/21/2025, with respect to the newly amended limitations have been fully considered but are not persuasive.
Applicant argues, beginning on Pg. 15 in the Remarks, that “Pouran's word representations and links between the words themselves would fail to disclose or suggest document representations and links between the documents.” Examiner respectfully disagrees. Examiner notes, Pouran in at least paragraphs [0017–0018, 0056, 0134], describes that their system generates document structures based on multiple sources of information using a graph transformer network because the GTN generates representations for heterogeneous document structure types from an original graph (i.e., an expert graph). Pouran clearly states its system creates these connections “including tasks where event trigger words and argument candidate words belong to different sentences in multiple documents” (Emphasis added). Pouran ¶0017. Pouran goes on in this regard stating “the improved network can extract arguments of event mentions over one or more documents to provide a complete view of information for events in these documents” (Emphasis added). Pouran 0018. See § 103 below for a detailed analysis.
Applicant argues, beginning on Pg. 16 in the Remarks, that “Pouran uses the adjacency matrix to determine the connections (e.g., the alleged links), but in contrast … while Pouran does describe graph transformer networks (GTNs), the cited portions of Pouran are directed to single documents and not directed to computing representations of documents based on an expert graph of the documents.” Examiner respectfully disagrees. Examiner notes that Pouran, in at least paragraphs [0017–0018, 0121, ] teaches that the GTN generates representations for heterogeneous document structure types including from multiple sources, for example from an original graph (i.e., an expert graph). Pouran teaches a deep neural network that “generates document structures based on multiple sources of information … As a result, the improved network can extract arguments of event mentions over one or more documents to provide a complete view of information for events in these documents” (Emphasis added). Pouran ¶¶0017–0018. See § 103 below for a detailed analysis.
Applicant argues, beginning on Pg. 16 in the Remarks, that “the Office suggests that Pouran's feed-forward neural network layer describes ‘using reinforcement learning.’” Specifically, Applicant argues that “the cited portion of Pouran merely mentioned that its GTN includes a decoder with a softmax layer, which a person skilled in the art would have understood is not a reinforcement learning module that uses reinforcement learning.” Examiner respectfully disagrees. Examiner contends that Pouran indeed teaches machine learning techniques described in the cited prior art section including reinforcement learning in at least paragraphs 0121 now explicitly cited below. See § 103 below for a detailed analysis. Furthermore, although not explicitly relied on for this rejection, Examiner notes that the secondary reference Mulligan teaches reinforcement learning in at least paragraph 0098.
Applicant argues, beginning on Pg. 16 in the Remarks, that “whereas the claims describe using reinforcement learning to select links that are presented to the application, which balances between exploration and exploitation as described above, Mulligan fails to perform such features and it would also not be ethically feasible to do so in the context of patient treatments, as described by Mulligan.” Examiner respectfully disagrees. Examiner points to Mulligan in at least paragraph 0098 which is reproduced and emphasized below:
In one aspect, generating and recommending the optimal medical actions 420 and/or matching the sections of existing CPGs 422, as described herein, may be performed using a wide variety of methods or combinations of methods, such as supervised learning, unsupervised learning, temporal difference learning, reinforcement learning and so forth. Some non-limiting examples of supervised learning which may be used with the present technology include AODE (averaged one-dependence estimators), artificial neural network, backpropagation, Bayesian statistics, naive bays classifier, Bayesian network, Bayesian knowledge base, case-based reasoning, decision trees, inductive logic programming, Gaussian process regression, gene expression programming, group method of data handling (GMDH), learning automata, learning vector quantization, minimum message length (decision trees, decision graphs, etc.), lazy learning, instance-based learning, nearest neighbor algorithm, analogical modeling, probably approximately correct (PAC) learning, ripple down rules, a knowledge acquisition methodology, symbolic machine learning algorithms, sub symbolic machine learning algorithms, support vector machines, random forests, ensembles of classifiers, bootstrap aggregating (bagging), boosting (meta-algorithm), ordinal classification, regression analysis, information fuzzy networks (IFN), statistical classification, linear classifiers, fisher's linear discriminant, logistic regression, perceptron, support vector machines, quadratic classifiers, k-nearest neighbor, hidden Markov models and boosting. Some non-limiting examples of unsupervised learning which may be used with the present technology include artificial neural network, data clustering, expectation-maximization, self-organizing map, radial basis function network, vector quantization, generative topographic map, information bottleneck method, IBSEAD (distributed autonomous entity systems based interaction), association rule learning, apriori algorithm, eclat algorithm, FP-growth algorithm, hierarchical clustering, single-linkage clustering, conceptual clustering, partitional clustering, k-means algorithm, fuzzy clustering, and reinforcement learning. Some non-limiting example of temporal difference learning may include Q-learning and learning automata. Specific details regarding any of the examples of supervised, unsupervised, temporal difference or other machine learning described in this paragraph are known and are within the scope of this disclosure. Also, when deploying one or more machine learning models, a computing device may be first tested in a controlled environment before being deployed in a public setting. Also even when deployed in a public environment (e.g., external to the controlled, testing environment), the computing devices may be monitored for compliance.
With respect to the remaining arguments including the dependent claims, without specific arguments detailing how the cited art does not teach each limitation, examiner maintains the rejections.
Thus, the § 103 rejections are maintained.
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 1–16 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 — Is the claim to a process, machine, manufacture, or composition of matter?
Yes, claim 1 is directed to a method i.e., a process.
Step 2A — Prong 1 Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
“generating a representation for each one of a plurality of heterogeneous documents contained in an expert graph having a plurality of links”
“computing a link score for each of the links of at least a first one of the documents based on the representations of the documents”
“selecting, for the first one of the documents, other ones of the documents as link targets based on the link scores”
These limitations, under their broadest reasonable interpretation, cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with the aid of pen and paper, a human can generate a representation for dissimilar documents, compute scores based on the representations, and select a document based on scores.
Step 2A — Prong 2 — Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
“computer-implemented method for generating accurate relationships among heterogeneous documents in a semantic graph for an application, the method comprising” — This limitation is reciting only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished such that it amounts no more than mere instructions to apply. See MPEP 2106.05(f); See also Electric Power Group, LLC v. Alstom, S.A., 830 F.3d 1350, 1356, 119 USPQ2d 1739 (Fed. Cir. 2016).
“using reinforcement learning” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f).
“forwarding the link targets to the application” — This limitation is insignificant extra-solution activity and is merely data outputting. See MPEP 2106.05(g).
Step 2B — Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception.
“forwarding the link targets to the application” — This limitation is directed to the activity of data outputting which is not an inventive concept because it is insignificant extra-solution activity of mere data gathering. See Mayo, 566 U.S. at 79, 101 USPQ2d at 1968; OIP Techs., Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1092-93 (Fed. Cir. 2015); MPEP 2106.05(g)(3). This limitation is well-understood, routine, and conventional because it involves transmitting information over a network. MPEP 2106.05(d)(II).
Regarding Claim 2:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 1 which included an abstract idea (see rejection for claim 1 above). This claim merely recites further limitations on the generating a representation limitation which is directed to abstract idea that can be performed in the human mind. The additional limitations:
“the expert graph is generated by applying a library of linking functions to the documents, each of the links linking a source document to a target document according to the linking functions” — This limitations is merely a continuation of the abstract idea in claim 1. Under their broadest reasonable interpretation, it covers mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with aid of pen of paper, a human can generate representations for data that has been generated using a library of functions.
“generating the representation for each one of the documents is performed by a document embedding model that includes a set of first parameters and is trained to generate the representations as an n-dimensional vector of real numbers based on raw data or text of the respective document, the expert graph and the set of first parameters” — This limitation is reciting only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished such that it amounts no more than mere instructions to apply. See MPEP 2106.05(f); See also Electric Power Group, LLC v. Alstom, S.A., 830 F.3d 1350, 1356, 119 USPQ2d 1739 (Fed. Cir. 2016). Examiner notes, though lengthy, this limitation merely generically recites a model to process data, based on the data and parameters, which amounts to no more than mere instructions to apply the abstract idea.
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding Claim 3:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 2 which included an abstract idea (see rejection for claim 2 above). This claim merely recites further limitations on the computing limitation which is directed to abstract idea that can be performed in the human mind. The additional limitations:
“wherein the link scores are computed by a link scoring model that includes a set of second parameters” — This limitations is merely a continuation of the abstract idea in claim 1. Under their broadest reasonable interpretation, it covers mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with aid of pen of paper, a human can compute link scores using a model for data that has been generated using a library of functions.
“wherein the link targets are selected using a reinforcement learning module that is trained using reinforcement learning to select the link targets based on the link scores and a link frequency such that the selected link targets are explorative and exploitable” — This limitation is reciting only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished such that it amounts no more than mere instructions to apply. See MPEP 2106.05(f); See also Electric Power Group, LLC v. Alstom, S.A., 830 F.3d 1350, 1356, 119 USPQ2d 1739 (Fed. Cir. 2016). Examiner notes, though lengthy, this limitation merely generically recites a trained model processing data, based on the data and parameters, which amounts to no more than mere instructions to apply the abstract idea.
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding Claim 4:
Step 1 — Is the claim to a process, machine, manufacture, or composition of matter?
Yes, claim 4 depends from claim 3 (see analysis of claim 3 above) which is directed to a method i.e., a process.
Step 2A — Prong 1 Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
“tracking browsing history with regard to the link targets”
These limitations, under their broadest reasonable interpretation, cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with the aid of pen and paper, a human can track browsing history. Examiner notes that a human could visually track browsing history based on the link targets.
Step 2A — Prong 2 — Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional limitation:
“training the document embedding model and the link scoring model using the browsing history as training data, to optimize the set of first parameters of the document embedding model and the set of second parameters of the link scoring model” — This limitation amounts to no more than mere instructions to apply the exception and is the equivalent to mere instruction to implement the abstract idea on a computer. See MPEP 2106.05(f). Training a model merely invokes computers or other machinery as a tool to perform an existing process.
Step 2B — Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception.
Regarding Claim 5:
Step 1 — Is the claim to a process, machine, manufacture, or composition of matter?
Yes, claim 5 depends from claim 4 (see analysis of claim 3 above) which is directed to a method i.e., a process.
Step 2A — Prong 1 Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
“wherein a reward is computed based on whether a respective target link is selected or ignored according to the browsing history”
These limitations, under their broadest reasonable interpretation, cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with the aid of pen and paper, a human can compute a reward based on the browsing history.
Step 2A — Prong 2 — Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional limitation:
“the reward is used to train the document embedding model and the link scoring model” — This limitation amounts to no more than mere instructions to apply the exception and is the equivalent to mere instruction to implement the abstract idea on a computer. See MPEP 2106.05(f). Training a model merely invokes computers or other machinery as a tool to perform an existing process.
Step 2B — Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception.
Regarding claim 6:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 4 which included an abstract idea (see rejection for claim 4 above). This claim merely recites a further limitation on the generating the representation limitation which is directed to an abstract idea that can be performed in the human mind. The additional limitation:
“wherein the document embedding model comprises a graph convolutional network (GCN)” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f).
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 7:
Step 1 — Is the claim to a process, machine, manufacture, or composition of matter?
Yes, claim 7 depends from claim 6 (see analysis of claim 6 above) which is directed to a method i.e., a process.
Step 2A — Prong 1 Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
“for the bottom layer, an initial embedding vector is computed for each respective document using a parameterized embedding function”
“for each higher layer, a next layer embedding vector is computed for each respective document by aggregating over a lower layer embedding vectors of neighboring documents in the expert graph”
These limitations, under their broadest reasonable interpretation, cover mathematical concepts (including mathematical relationships, mathematical formulas or equations, mathematical calculations) See MPEP 2106.04(a)(2). In particular, the above initial embedding vector is computed and next layer embedding vector is computed are mathematical concepts directed to computing (i.e., calculating) mathematical equations and are therefore abstract ideas (see present disclosure paragraphs 0036–0037).
Step 2A — Prong 2 — Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional limitation:
“wherein the GCN includes a bottom layer and one or more higher layers” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f).
Step 2B — Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception.
Regarding Claim 8:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 7 which included an abstract idea (see rejection for claim 7 above). This claim merely recites a further limitation on the next layer embedding vector is computed limitation which is directed to an abstract idea that covers mathematical concepts. The additional limitations:
“aggregating over the lower layer embedding vectors of neighboring documents for each respective linking function” — These limitations are merely a continuation of the abstract idea in claim 7. Under their broadest reasonable interpretation, they cover mathematical concepts (including mathematical relationships, mathematical formulas or equations, mathematical calculations). See MPEP 2106.04(a)(2). In particular, the above “aggregating” limitation corresponds to mathematical relationships and mathematical calculations (see present disclosure paragraphs 0036–0037). See MPEP 2106.04(a)(2).
“aggregating over all linking functions in the library of linking functions” — These limitations are merely a continuation of the abstract idea in claim 7. Under their broadest reasonable interpretation, they cover mathematical concepts (including mathematical relationships, mathematical formulas or equations, mathematical calculations). See MPEP 2106.04(a)(2). In particular, the above “aggregating” limitation corresponds to mathematical relationships and mathematical calculations (see present disclosure paragraphs 0036–0037). See MPEP 2106.04(a)(2).
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 9:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 4 which included an abstract idea (see rejection for claim 4 above). This claim merely recites a further limitation on the computing a link score limitation which is directed to an abstract idea that can be performed in the human mind. The additional limitation:
“wherein the link scoring model comprises a neural network” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f).
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 10:
Step 1 — Is the claim to a process, machine, manufacture, or composition of matter?
Yes, claim 10 depends from claim 1 (see analysis of claim 6 above) which is directed to a method i.e., a process.
Step 2A — Prong 1 Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
“computing a respective link frequency for each respective link”
“wherein the link targets are selected further based on the link frequencies in addition to the link scores” — Examiner notes this limitation is merely a continuation of the “selecting” abstract idea in claim 1.
These limitations, under their broadest reasonable interpretation, cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with the aid of pen and paper, a human can compute link frequencies and select documents based on link scores.
Step 2A — Prong 2 — Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that 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, there are no additional elements that amount to significantly more than the judicial exception.
Regarding Claim 11:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 10 which included an abstract idea (see rejection for claim 10 above). This claim merely recites a further limitation on the computing a link score and computing a link frequency limitations, respectively, which are directed to an abstract idea that can be performed in the human mind. The additional limitations:
“wherein each link score indicates a degree of relevance of the first one of the documents and a respective link target, and each link frequency indicates a frequency of the respective link target being suggested” — These limitations are merely a continuation of the abstract ideas in claims 1 and 10. Under their broadest reasonable interpretation, they cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with aid of pen of paper, a human can compute a link score based on a degree of relevance and compute a link frequency for specific links.
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding Claim 12:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 11 which included an abstract idea (see rejection for claim 11 above). This claim merely recites a further limitation on the selecting limitation which is directed to an abstract idea that can be performed in the human mind. The additional limitations:
“wherein the link targets are selected by optimizing a function that is proportional to the link score and inversely proportional to the link frequency” — This limitation is merely a continuation of the abstract idea in claim 1. Under its broadest reasonable interpretation, it covers mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with aid of pen of paper, a human can make a selection based on an optimized function.
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding claim 13:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 1 which included an abstract idea (see rejection for claim 1 above). This claim merely recites further limitations which are directed to mere instructions to apply the abstract ideas. Examiner notes these limitations effectively repeat steps of the method. The additional limitations:
“accommodating a new document by: applying the expert linking functions to the new document to extract new links between the new document and the documents” — This limitation amounts to no more than mere instructions to apply the exception and is the equivalent to mere instruction to implement the abstract idea on a computer. See MPEP 2106.05(f). Repeating steps to process new data merely invokes computers or other machinery as a tool to perform an existing process.
“generating a new representation for the new document based on the representations of the documents that are stored in cache” — This limitation is reciting only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished such that it amounts no more than mere instructions to apply. See MPEP 2106.05(f); See also Electric Power Group, LLC v. Alstom, S.A., 830 F.3d 1350, 1356, 119 USPQ2d 1739 (Fed. Cir. 2016).
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Regarding Claim 14:
Step 1 — Is the claim to a process, machine, manufacture, or composition of matter?
Yes, claim 1 is directed to a system i.e., a machine.
Step 2A — Prong 1 Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
“generating a representation for each one of a plurality of heterogeneous documents contained in an expert graph having a plurality of links”
“computing a link score for each of the links of at least a first one of the documents based on the representations of the documents”
“selecting, for the first one of the documents, other ones of the documents as link targets based on the link scores”
These limitations, under their broadest reasonable interpretation, cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with the aid of pen and paper, a human can generate a representation for dissimilar documents, compute scores based on the representations, and select a document based on scores.
Step 2A — Prong 2 — Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
“A system for generating accurate relationships among heterogeneous documents in a semantic graph for an application” — This limitation is reciting only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished such that it amounts no more than mere instructions to apply. See MPEP 2106.05(f); See also Electric Power Group, LLC v. Alstom, S.A., 830 F.3d 1350, 1356, 119 USPQ2d 1739 (Fed. Cir. 2016).
“the system comprising one or more hardware processors which, alone or in combination, are configured to provide for execution of the following steps” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f)
“using reinforcement learning” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f).
“forwarding the link targets to the application” — This limitation is insignificant extra-solution activity and is merely data outputting. See MPEP 2106.05(g).
Step 2B — Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception.
“forwarding the link targets to the application” — This limitation is directed to the activity of data outputting which is not an inventive concept because it is insignificant extra-solution activity of mere data gathering. See Mayo, 566 U.S. at 79, 101 USPQ2d at 1968; OIP Techs., Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1092-93 (Fed. Cir. 2015); MPEP 2106.05(g)(3). This limitation is well-understood, routine, and conventional because it involves transmitting information over a network. MPEP 2106.05(d)(II).
Regarding Claim 15:
Step 1 — Is the claim to a process, machine, manufacture, or composition of matter?
Yes, claim 1 is directed to a system i.e., a machine.
Step 2A — Prong 1 Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Yes, the claim recites an abstract idea.
“generating a representation for each one of a plurality of heterogeneous documents contained in an expert graph having a plurality of links”
“computing a link score for each of the links of at least a first one of the documents based on the representations of the documents”
“selecting, for the first one of the documents, other ones of the documents as link targets based on the link scores”
These limitations, under their broadest reasonable interpretation, cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with the aid of pen and paper, a human can generate a representation for dissimilar documents, compute scores based on the representations, and select a document based on scores.
Step 2A — Prong 2 — Does the claim recite additional elements that integrate the judicial exception into a practical application?
No, there are no additional elements that integrate the judicial exception into a practical application. The additional elements:
“A tangible, non-transitory computer-readable medium having instructions thereon, which upon execution by one or more processors, alone or in combination, provide for execution of a method comprising” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f)
“using reinforcement learning” — This limitation is reciting generic computer components at a high-level of generality (i.e., as a generic computer component performing a generic computer function) such that it amounts no more than mere instructions to apply the exception using a generic computer component. See MPEP 2106.05(f).
“forwarding the link targets to the application” — This limitation is insignificant extra-solution activity and is merely data outputting. See MPEP 2106.05(g).
Step 2B — Does the claim recite additional elements that amount to significantly more than the judicial exception?
No, there are no additional elements that amount to significantly more than the judicial exception.
“forwarding the link targets to the application” — This limitation is directed to the activity of data outputting which is not an inventive concept because it is insignificant extra-solution activity of mere data gathering. See Mayo, 566 U.S. at 79, 101 USPQ2d at 1968; OIP Techs., Inc. v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1092-93 (Fed. Cir. 2015); MPEP 2106.05(g)(3). This limitation is well-understood, routine, and conventional because it involves transmitting information over a network. MPEP 2106.05(d)(II).
Regarding Claim 16:
The claim is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim is dependent on claim 1 which included an abstract idea (see rejection for claim 1 above). This claim merely recites a further limitation on the plurality of links limitation which is directed to an abstract idea that can be performed in the human mind. The additional limitations:
“wherein each of the plurality of links is an ordered triple indicating a source document, a target document, and an expert linker of a plurality of expert linkers that generated the link between the source document and the target document” — These limitations are merely a continuation of the abstract idea in claim 1. Under their broadest reasonable interpretation, they cover mental processes, concepts performed in the human mind (including an observation, evaluation, judgment, opinion). See MPEP 2106.04(a)(2). In particular, with aid of pen of paper, a human can generate a representation for dissimilar documents where the representations are generated output in the form of a triple.
Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d)I.), failing Step 2A Prong 2. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1–3, 10–12, and 14–16 are rejected under 35 U.S.C. 103 as being unpatentable over Pouran Ben Veyseh et al., (US 20220318505 A1), hereinafter “Pouran”, in view of Mulligan et al., (US 20210057098 A1), hereinafter “Mulligan”.
Regarding claim 1, Pouran teaches:
a computer-implemented method for generating accurate relationships among heterogeneous documents in a semantic graph for an application, the method comprising (Pouran ¶0005: “method, apparatus, and non-transitory computer readable medium for natural language processing … generating a plurality of document structures including a semantic structure for the document based on the word representation vectors, a syntax structure representing dependency relationships between the words”; see also Pouran ¶0054: “In some cases, deep learning models can operate directly over raw input data such as text or images, and enable connections between a set of entities, and a user can interpret those connections after a knowledge accumulation phase and make inferences based on prior knowledge”):
generating a representation for each one of a plurality of heterogeneous documents contained in an expert graph having a plurality of links (Pouran ¶¶0017–0018, ¶0056, ¶0121: “A Graph Transformer Network (GTN) is capable of generating new graph structures, which involve identifying useful connections between unconnected nodes on the original graph, while learning effective node representation on the new graphs in an end-to-end fashion. Graph Transformer layer, a core layer of the GTN, learns a soft selection of edge types and composite relations for generating useful multi-hop connections or meta-paths. In some cases, GTNs learn new graph structures, based on data and tasks without domain knowledge, and can yield powerful node representation via convolution on the new graphs. Without domain-specific graph preprocessing, GTNs outperform existing technology that require pre-defined meta-paths from domain knowledge”; see also Pouran ¶0134: “The event argument extraction network is a multi-hop reasoning for event argument extractor with heterogeneous document structure types”—[wherein the GTN generates representations for heterogeneous document structure types from an original graph (i.e., an expert graph)]);
computing a link score for each of the links of at least a first one of the documents based on the representations of the documents (Pouran ¶0068: “Document structures 420 are used to facilitate connection and reasoning between important context words for prediction. In some cases, document structures 420 may refer to interaction graphs, utilizing designated objects in one or more documents (e.g., words, entity mentions, sentences) to form nodes, and different information sources and heuristics to establish the edges. For example, document structures 420 may be represented using adjacency matrices for which the value or score at one cell indicates the importance of a node toward the other one for representation learning and role prediction in event argument extraction. Such direct connections between nodes enable objects, which are sequentially far from each other in the documents, to interact and produce useful information”—[wherein the value or score (i.e., computed link score) indicates (i.e., represents) the similarity between objects (i.e., documents) based on the document structures (i.e., representations of the documents)]);
selecting, for the first one of the documents, other ones of the documents as link targets based on the link scores using reinforcement learning (Pouran ¶0056: “Graph Transformer layer, a core layer of the GTN, learns a soft selection of edge types and composite relations for generating useful multi-hop connections or meta-paths”; see also Pouran ¶0057–0058: “In some examples, decoder 335 applies a feed-forward neural network to the relationship representation vector. In some examples, decoder 335 applies a softmax layer to an output of the feed-forward neural network, where the relationship is predicted based on the output of the softmax layer. In some examples, the decoder 335 includes a feed-forward layer and a softmax layer. Decoder 335 is an example of, or includes aspects of, the corresponding element described with reference to FIG. 4. A softmax function is used as the activation function of the neural network to normalize the output of the network to a probability distribution over predicted output classes. After applying the softmax function, each component of the feature map is in the interval (0, 1) and the components add up to one. These values are interpreted as probabilities”; see also Pouran ¶0121: “unsupervised learning is one of three basic machine learning paradigms, alongside unsupervised learning and reinforcement learning.”—[(emphasis added) wherein the system teaches using its methods alongside (i.e., based on) reinforcement learning techniques]).
Pouran does not appear to explicitly teach:
forwarding the link targets to the application.
However, Mulligan teaches:
forwarding the link targets to the application (Mulligan ¶0116–0117: “The matching module 624 may compare the data and provide a sorted list of sections of clinical guidelines matching the current patient profile 622. That is, the matching module 624 extracts a patient pathway from the current patient profile 622. For each patient pathway “H”, the current patient profile 622 uses the patient pathways model M1 to build a vector representation of patient pathway H. The matching module 624 may determine and/or compute one or more similarity metrics between the vector representation of the patient pathway H and vector representations of the guidelines in clinical guideline model M2. The similarity metric provides a sorted list of sections of guidelines/CPGs that are closest to the patient pathway H. The sections of guidelines/CPGs that are the closest/most similar are the most appropriate/determined section for patient pathway H … That is, the feedback module 628 may provide a structured database of feedbacks … The feedback data 630 may also be communicated back to the clinical guidelines features learning module 616 and to the matching module 624”—[wherein the matching module provides the list of sections of guidelines/CPGs (i.e., link targets) to the clinical guidelines features learning module and the matching module (i.e., to an application)]).
The methods of Pouran, the teachings of Mulligan, and the instant application are analogous art because they pertain to incorporating machine learning techniques to link heterogenous data.
It would be obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the methods of Pouran with the teachings of Mulligan to provide a mechanism to explicitly transfer the learned correlated pathways and connections to other applications for use. One would be motivated to do so to allow the system to improve the rankings and to allow users to provide feedback on the matching thus improving the rankings (Mulligan ¶0116–0117: “The sections of guidelines/CPGs that are the closest/most similar are the most appropriate/determined section for patient pathway H. If available, the present invention may use the feedback data 630 collected from the domain expert(s) 632 via a feedback module 628 to improve the ranking in the sorted list … The feedback module 628 may receive as input data from one or more domain experts such as, for example, the domain expert 632, and provide feedback data 630 on the output produced by the matching module 624. That is, the feedback module 628 may provide a structured database of feedbacks. The feedback module 628 allows for users to provide feedback about the matching sections of CPGs using one or more selected/defined user interfaces of a computing system”).
Regarding claim 2, Pouran in view of Mulligan teaches all the limitations of claim 1.
Pouran teaches:
the expert graph is generated by applying a library of linking functions to the documents, each of the links linking a source document to a target document according to the linking functions (Pouran ¶0052: “One or more embodiments of the present disclosure combine different information sources to generate effective document structures for event argument extraction tasks. The event argument extraction network 315 produces document structures based on knowledge from syntax (i.e., dependency trees), discourse (i.e., coreference links), and semantic similarity. Semantic similarity depends on contextualized representation vectors to compute interaction scores between nodes and relies on using external knowledge bases to enrich document structures for event argument extraction. The words in the documents are linked to the entries in one or more external knowledge bases and exploit the entry similarity in knowledge bases to obtain word similarity scores for the structures. In some examples, lexical database (e.g., WordNet) is used as the knowledge base and tools (e.g., word sense disambiguation or WSD) are applied to facilitate word-entry linking. The linked entry or node in WordNet can provide expert knowledge on the meanings of the words (e.g., glossary and hierarchy information). Such expert knowledge complements the contextual information of words and enhances the semantic-based document structures for event argument extraction. In some embodiments, the event argument extraction network 315 uses one or more external knowledge bases for document structures in information extraction tasks”—[(emphasis added) wherein the system uses external sources (e.g., expert external knowledge bases) to link the documents based on dependency trees, discourse, and semantic similarity which all use different functions to find matches]); and
generating the representation for each one of the documents is performed by a document embedding model that includes a set of first parameters and is trained to generate the representations as an n-dimensional vector of real numbers based on raw data or text of the respective document, the expert graph and the set of first parameters (Pouran ¶0068: “In an embodiment, the structure component 415 can generate document structures 420 based on syntactic dependency trees of a set of sentences, considering discourse structures (e.g., coreference links) to induce document structures 420, and employing semantic representation-based similarity between tokens to infer document structures 420”; see also Pouran ¶0072: “the document structures 420 are input to the relationship encoder 440 to produce relationship representation vector 445. The relationship encoder 440 is an example of, or includes aspects of, the corresponding element described with reference to FIG. 3. The resulting combined structures are used to learn representation vectors for event argument extraction based on graph convolutional networks (GCNs), which are convolutional neural networks used to encode graphs”—[wherein the relationship encoder uses a trained neural network model (e.g., graph convolutional network) to produce relationship vector 445 (i.e., an n-dimensional vector of real numbers) based on the document data, the original graph, and parameters (e.g., syntactic dependency trees of a set of sentences, considering discourse structures, and semantic similarity)]).
Regarding claim 3, Pouran in view of Mulligan teaches all the limitations of claim 2.
Pouran teaches:
wherein the link scores are computed by a link scoring model that includes a set of second parameters, and (Pouran ¶0092–0093: “In some examples, the syntax-based document structure is based on sentence-level event argument extraction where dependency parsing trees of input sentences reveal important context, i.e., via the shortest dependency paths to connect event triggers and arguments, and guide the interaction modeling between words for argument role prediction by LSTM cells. The dependency trees for the sentences in D are used to provide information for the document structures for event argument extraction. In some embodiments, the dependency relations or connections between pairs of words in W are leveraged to compute interaction scores”—[wherein the system uses LSTM cells (i.e., link scoring model) and the dependency relations to compute interaction scores (i.e., compute link scores with a set of second parameters)]).
Pouran does not appear to explicitly teach:
wherein the link targets are selected using a reinforcement learning module that is trained using reinforcement learning to select the link targets based on the link scores and a link frequency such that the selected link targets are explorative and exploitable.
However, Mulligan teaches:
wherein the link targets are selected using a reinforcement learning module that is trained using reinforcement learning to select the link targets based on the link scores and a link frequency such that the selected link targets are explorative and exploitable (Mulligan ¶0092: “In one aspect, the optimal medical actions 420 and the one or more sections of existing CPGs 422 may be text based and text fragments may be converted to term frequency-inverse document frequency (“TF-IDF”) vectors where a cosine similarity between the optimal medical actions 420 and the one or more sections of existing CPGs 422 is determined. The cosine similarity among vectors may be used to rank the optimal medical actions 420 and the one or more sections of existing CPGs 422. A predetermined threshold may be used to determine and establish when the optimal medical actions 420 match the one or more sections of existing CPGs 422 (e.g., matching occurs if a matching score exceeds a predetermined threshold)”; see also Mulligan ¶0098: “In one aspect, generating and recommending the optimal medical actions 420 and/or matching the sections of existing CPGs 422, as described herein, may be performed using a wide variety of methods or combinations of methods, such as supervised learning, unsupervised learning, temporal difference learning, reinforcement learning and so forth”—[wherein the system utilizes reinforcement learning to perform the method of converting the data to term Frequency-inverse document frequency vectors (i.e., link frequency) used to optimize the cosine similarity (i.e., the link score) in order to rank (i.e., select) the data for further recommendations (i.e., explorative and exploitable)]).
The same motivation that was utilized for combining Pouran with Mulligan, as set forth in claim 1, is equally applicable to claim 3.
Regarding claim 10, Pouran in view of Mulligan teaches all the limitations of claim 1.
Mulligan teaches:
computing a respective link frequency for each respective link (Mulligan ¶0092: “In one aspect, the optimal medical actions 420 and the one or more sections of existing CPGs 422 may be text based and text fragments may be converted to term frequency-inverse document frequency (“TF-IDF”) vectors where a cosine similarity between the optimal medical actions 420 and the one or more sections of existing CPGs 422 is determined. The cosine similarity among vectors may be used to rank the optimal medical actions 420 and the one or more sections of existing CPGs 422. A predetermined threshold may be used to determine and establish when the optimal medical actions 420 match the one or more sections of existing CPGs 422 (e.g., matching occurs if a matching score exceeds a predetermined threshold)”);
wherein the link targets are selected further based on the link frequencies in addition to the link scores (Mulligan ¶0092: “In one aspect, the optimal medical actions 420 and the one or more sections of existing CPGs 422 may be text based and text fragments may be converted to term frequency-inverse document frequency (“TF-IDF”) vectors where a cosine similarity between the optimal medical actions 420 and the one or more sections of existing CPGs 422 is determined. The cosine similarity among vectors may be used to rank the optimal medical actions 420 and the one or more sections of existing CPGs 422. A predetermined threshold may be used to determine and establish when the optimal medical actions 420 match the one or more sections of existing CPGs 422 (e.g., matching occurs if a matching score exceeds a predetermined threshold)”; see also Mulligan ¶0098: “In one aspect, generating and recommending the optimal medical actions 420 and/or matching the sections of existing CPGs 422, as described herein, may be performed using a wide variety of methods or combinations of methods, such as supervised learning, unsupervised learning, temporal difference learning, reinforcement learning and so forth”—[wherein the system converts the data to term Frequency-inverse document frequency vectors (i.e., link frequency) used to optimize the cosine similarity (i.e., the link score) in order to rank (i.e., select) the data for further recommendations)).
The same motivation that was utilized for combining Pouran with Mulligan, as set forth in claim 1, is equally applicable to claim 10.
Regarding claim 11, Pouran in view of Mulligan teaches all the limitations of claim 10.
Pouran teaches:
wherein each link score indicates a degree of relevance of the first one of the documents and a respective link target (Pouran ¶0068: “Document structures 420 are used to facilitate connection and reasoning between important context words for prediction. In some cases, document structures 420 may refer to interaction graphs, utilizing designated objects in one or more documents (e.g., words, entity mentions, sentences) to form nodes, and different information sources and heuristics to establish the edges. For example, document structures 420 may be represented using adjacency matrices for which the value or score at one cell indicates the importance of a node toward the other one for representation learning and role prediction in event argument extraction. Such direct connections between nodes enable objects, which are sequentially far from each other in the documents, to interact and produce useful information”—[wherein the value or score (i.e., computed link score) indicates (i.e., represents) the similarity between objects (i.e., documents) based on the document structures (i.e., representations of the documents)), and
Mulligan teaches:
each link frequency indicates a frequency of the respective link target being suggested (Mulligan ¶0092: “In one aspect, the optimal medical actions 420 and the one or more sections of existing CPGs 422 may be text based and text fragments may be converted to term frequency-inverse document frequency (“TF-IDF”) vectors where a cosine similarity between the optimal medical actions 420 and the one or more sections of existing CPGs 422 is determined. The cosine similarity among vectors may be used to rank the optimal medical actions 420 and the one or more sections of existing CPGs 422. A predetermined threshold may be used to determine and establish when the optimal medical actions 420 match the one or more sections of existing CPGs 422 (e.g., matching occurs if a matching score exceeds a predetermined threshold)”; see also Mulligan ¶0098: “In one aspect, generating and recommending the optimal medical actions 420 and/or matching the sections of existing CPGs 422, as described herein, may be performed using a wide variety of methods or combinations of methods, such as supervised learning, unsupervised learning, temporal difference learning, reinforcement learning and so forth”—[wherein the system converts the data to term Frequency-inverse document frequency vectors (i.e., link frequency) used to optimize the cosine similarity (i.e., the link score) in order to rank (i.e., select) the data for further recommendations)).
The same motivation that was utilized for combining Pouran with Mulligan, as set forth in claim 1, is equally applicable to claim 11.
Regarding claim 12, Pouran in view of Mulligan teaches all the limitations of claim 11.
Pouran teaches:
wherein each link score indicates a degree of relevance of the first one of the documents and a respective link target (Pouran ¶0068: “Document structures 420 are used to facilitate connection and reasoning between important context words for prediction. In some cases, document structures 420 may refer to interaction graphs, utilizing designated objects in one or more documents (e.g., words, entity mentions, sentences) to form nodes, and different information sources and heuristics to establish the edges. For example, document structures 420 may be represented using adjacency matrices for which the value or score at one cell indicates the importance of a node toward the other one for representation learning and role prediction in event argument extraction. Such direct connections between nodes enable objects, which are sequentially far from each other in the documents, to interact and produce useful information”—[wherein the value or score (i.e., computed link score) indicates (i.e., represents) the similarity between objects (i.e., documents) based on the document structures (i.e., representations of the documents) to make a prediction (i.e., target)), and
Mulligan teaches:
each link frequency indicates a frequency of the respective link target being suggested (Mulligan ¶0092: “In one aspect, the optimal medical actions 420 and the one or more sections of existing CPGs 422 may be text based and text fragments may be converted to term frequency-inverse document frequency (“TF-IDF”) vectors where a cosine similarity between the optimal medical actions 420 and the one or more sections of existing CPGs 422 is determined. The cosine similarity among vectors may be used to rank the optimal medical actions 420 and the one or more sections of existing CPGs 422. A predetermined threshold may be used to determine and establish when the optimal medical actions 420 match the one or more sections of existing CPGs 422 (e.g., matching occurs if a matching score exceeds a predetermined threshold)”; see also Mulligan ¶0098: “In one aspect, generating and recommending the optimal medical actions 420 and/or matching the sections of existing CPGs 422, as described herein, may be performed using a wide variety of methods or combinations of methods, such as supervised learning, unsupervised learning, temporal difference learning, reinforcement learning and so forth”—[wherein the system converts the data to term Frequency-inverse document frequency vectors (i.e., link frequency) used to optimize the cosine similarity (i.e., the link score) in order to rank (i.e., select) the data (i.e., target) for further recommendations)).
The same motivation that was utilized for combining Pouran with Mulligan, as set forth in claim 1, is equally applicable to claim 12.
Regarding claim 14, Pouran teaches:
a system for generating accurate relationships among heterogeneous documents in a semantic graph for an application, the system comprising one or more hardware processors which, alone or in combination, are configured to provide for execution of the following steps: (Pouran ¶0005: “method, apparatus, and non-transitory computer readable medium for natural language processing … generating a plurality of document structures including a semantic structure for the document based on the word representation vectors, a syntax structure representing dependency relationships between the words”; see also Pouran ¶0054: “In some cases, deep learning models can operate directly over raw input data such as text or images, and enable connections between a set of entities, and a user can interpret those connections after a knowledge accumulation phase and make inferences based on prior knowledge”; see also Pouran ¶0120: “FIG. 6 shows an example of a process for training an event argument extraction network according to aspects of the present disclosure. In some examples, these operations are performed by a system including a processor executing a set of codes to control functional elements of an apparatus”).
Regarding claim 15, Pouran teaches:
a tangible, non-transitory computer-readable medium having instructions thereon, which upon execution by one or more processors, alone or in combination, provide for execution of a method comprising: (Pouran ¶0005: “method, apparatus, and non-transitory computer readable medium for natural language processing … generating a plurality of document structures including a semantic structure for the document based on the word representation vectors, a syntax structure representing dependency relationships between the words”; see also Pouran ¶0054: “In some cases, deep learning models can operate directly over raw input data such as text or images, and enable connections between a set of entities, and a user can interpret those connections after a knowledge accumulation phase and make inferences based on prior knowledge”; see also Pouran ¶0120: “FIG. 6 shows an example of a process for training an event argument extraction network according to aspects of the present disclosure. In some examples, these operations are performed by a system including a processor executing a set of codes to control functional elements of an apparatus”).
Regarding the remaining limitation in claims 14–15, although varying in scope, the limitations of claims 14–15 are substantially the same as the limitations of claim 1, respectively. Thus, the remaining limitations in claims 14–15 are rejected using the same reasoning and analysis as claim 1 above, respectively.
Regarding claim 16, Pouran in view of Mulligan teaches all the limitations of claim 1.
Pouran teaches:
wherein each of the plurality of links is an ordered triple (Pouran ¶0121, ¶0137: “Supervised learning is a machine learning technique based on learning a function that maps an input to an output based on example input-output pairs. Supervised learning generates a function for predicting labeled data based on labeled training data consisting of a set of training examples. In some cases, each example is a pair consisting of an input object (typically a vector) and a desired output value (i.e., a single value, or an output vector” and “The development set of the RAMS dataset is used to fine-tune the hyper-parameters of the network model. After the fine-tuning process, the hyper-parameters used for a dataset (i.e., BNB) are 1e-5 for learning rate of the Adam optimizer, 32 for the mini-batch size, 30 dimensions for the position embeddings, 200 hidden units for the feed-forward network, a bidirectional LSTM (Bi-LSTM) and GCN layers, 2 layers for Bi-LSTM and GCN models (G=2), and C=3 channels for GTN with M=3 intermediate structures in each channel”—[wherein the BRI of an ordered triple is a set of three numbers or values arranged in a specific order, and wherein the machine learning input matches the output which includes C=3 channels for GTN M=3 intermediate structures in each of the three channels]),
indicating a source document, a target document, and an expert linker of a plurality of expert linkers that generated the link between the source document and the target document (Pouran ¶¶0017–0018: “One or more embodiments of the present disclosure provide an improved event argument extraction apparatus that can perform document-level event argument extraction tasks, including tasks where event trigger words and argument candidate words belong to different sentences in multiple documents. In some examples, the event argument extraction network includes a deep neural network that generates document structures based on multiple sources of information such as syntax, semantic and discourse. According to one embodiment, a graph transformer network (GTN) is used to combine these document structures … By applying the unconventional step of generating multiple document structures, one or more embodiments of the present disclosure provide an event argument extraction network that can perform efficient event argument extraction at a document level. The improved network is scalable to scenarios where an event trigger word and an argument candidate word are located far from each other in different sentences or documents … As a result, the improved network can extract arguments of event mentions over one or more documents to provide a complete view of information for events in these documents)”—[(emphasis added) wherein the system generates document structures based on multiple sources of information using a graph transformer network to combine the structures into an output including C=3 channels for GTN with M=3 intermediate structures in each channel]).
Claims 4–9 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Pouran in view of Mulligan and further in view of Gutierrez et al., (US 20220309037 A1), hereinafter “Gutierrez”.
Regarding claim 4, Pouran in view of Mulligan teaches all the limitations of claim 3.
Pouran in view of Mulligan does not appear to explicitly teach:
tracking browsing history with regard to the link targets; and
training the document embedding model and the link scoring model using the browsing history as training data, to optimize the set of first parameters of the document embedding model and the set of second parameters of the link scoring model.
However, Gutierrez teaches:
tracking browsing history with regard to the link targets (Gutierrez ¶0623: “Identifying relevant information and/or prepopulating input fields are not limited to saving or to the depicted sections. For instance, the analytics server may use the methods and systems described herein to identify and/or generate information associated with any of the metadata fields or features depicted in FIG. 54 (not just the summary section 5402). The analytics server can automatically collect data associated with a variety of fields, such as the ones shown in the overlay 5400. As described herein, the analytics server can automatically generate and revise (update and/or recommend) the nodal data structure using data captured as one or more users are accessing or otherwise interacting with data (e.g., working or browsing the internet), through, for example, manual saving, general browsing, or through predetermined rules that only record certain types of data while working normally (e.g., to processing overhead down and/or avoid the nodal data structure growing in size too much)”); and
training the document embedding model and the link scoring model using the browsing history as training data, to optimize the set of first parameters of the document embedding model and the set of second parameters of the link scoring model (Gutierrez ¶0624: “While the user is operating his/her computer to access data (e.g., writing emails, working on files, browsing the Internet, etc.), the analytics server can continuously or periodically monitor and/or analyze the user's computer and related electronic context to revise the nodal data structure. The analytics server can also retroactively review recordings of user activity, screen recordings, network traffic recordings, or other historical data. For instance, the analytics server may retrieve the user's recorded history and any corresponding metadata according to the user's configured settings, and augment or modify it as specified (e.g., with standard metadata about the user's location, device, and other attributes discussed herein). The analytics server may also classify what type of data is being monitored/retrieved. Based on the collected data, the analytics server may also search for additional metadata for a given historical record from third-party sources (e.g., find a URL in a user's browsing history and access that website remotely to gather additional metadata). The analytics server may then interrelate the collected data to other related data within the nodal data structure. As a result, the analytics server may generate/revise relationships between what the user has historically done (e.g., electronic content accessed by the user) and other content within the nodal data structure”—[wherein the analytics server uses the tracked browsing history to generate/revise the nodal data structure to identify relevant information]).
The methods of Pouran, the teachings of Gutierrez, and the instant application are analogous art because they pertain to using machine learning techniques to organize and find relationships in data.
It would be obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the methods of Pouran with the teachings of Gutierrez to provide additional data gathering resources. One would be motivated to do so to aid in detecting related electronic context and improve established relationships (Gutierrez ¶0625: “Using this data, the analytics server may revise the nodal data structure and improve the established relationships between nodes”).
Regarding claim 5, Pouran in view of Mulligan and Gutierrez teaches all the limitations of claim 4.
Gutierrez teaches:
wherein a reward is computed based on whether a respective target link is selected or ignored according to the browsing history, and the reward is used to train the document embedding model and the link scoring model (Gutierrez ¶0167: “In order to identify whether two nodes are related, the analytics server may first generate an initial label that includes data indicating that two files may be related (also referred to herein as implicit dataset or implicit feedback dataset). The analytics server may then use a file correlation algorithm and/or scoring algorithm to generate a score that represents a distance between two nodes. In order to achieve this, the analytics server may execute various scoring algorithms and or AI/ML models”; see also Gutierrez ¶0624: “For instance, the analytics server may retrieve the user's recorded history and any corresponding metadata according to the user's configured settings, and augment or modify it as specified (e.g., with standard metadata about the user's location, device, and other attributes discussed herein). The analytics server may also classify what type of data is being monitored/retrieved. Based on the collected data, the analytics server may also search for additional metadata for a given historical record from third-party sources (e.g., find a URL in a user's browsing history and access that website remotely to gather additional metadata). The analytics server may then interrelate the collected data to other related data within the nodal data structure. As a result, the analytics server may generate/revise relationships between what the user has historically done (e.g., electronic content accessed by the user) and other content within the nodal data structure”; see also Gutierrez ¶0213: “Referring back to FIG. 8, at step 830, the analytics server may periodically reclassify and relabel nodes to achieve better results. The analytics server improves the quality of initial labeling and the AI/ML powered scoring by considering user behavior. For example, the analytics server may provide an option for users to accept and/or reject recommendations and have their actions influence the earlier labels and models described herein”—[wherein the analytics server interrelates the collected browser data with the nodal structure using the score (i.e., computed reward) by reclassifying and relabeling nodes (i.e., training the model)]).
The same motivation that was utilized for combining Pouran with Gutierrez, as set forth in claim 4, is equally applicable to claim 5.
Regarding claim 6, Pouran in view of Mulligan and Gutierrez teaches all the limitations of claim 4.
Pouran teaches:
wherein the document embedding model comprises a graph convolutional network (GCN) (Pouran ¶0038: “In some examples, the document encoder comprises a word encoder, a position embedding table, and an LSTM. In some examples, the structure component comprises a lexical database, a dependency parser, and a coreference network. In some examples, the relationship encoder comprises a GTN and a graph convolution network (GCN). In some examples, the decoder comprises a feed-forward layer and a softmax layer”).
Regarding claim 7, Pouran in view of Mulligan and Gutierrez teaches all the limitations of claim 6.
Pouran teaches:
wherein the GCN includes a bottom layer and one or more higher layers (Pouran ¶0038: “In some examples, the document encoder comprises a word encoder, a position embedding table, and an LSTM. In some examples, the structure component comprises a lexical database, a dependency parser, and a coreference network. In some examples, the relationship encoder comprises a GTN and a graph convolution network (GCN). In some examples, the decoder comprises a feed-forward layer and a softmax layer”), and
wherein: for the bottom layer, an initial embedding vector is computed for each respective document using a parameterized embedding function (Pouran ¶0111: “As such, each of these GCN models involves G layers that produce the hidden vectors h 1 k,t, . . . , h N k,t at the t-th layer of the k-th GCN model for the words in D (1≤k≤C, 1≤t<G) … where Uk,t is the weight matrix for the t-th layer of the k-th GCN model and the input vectors h i k,0 for the GCN models are obtained from the contextualized representation vectors H (i.e., h i k,0=hi for all 1≤k≤C, 1≤i≤N)”); and
for each higher layer, a next layer embedding vector is computed for each respective document by aggregating over a lower layer embedding vectors of neighboring documents in the expert graph (Pouran ¶0113: “Next, the hidden vectors in the last layers of all the GCN models (at the G-th layers) for wi (i.e., h i 1,G, h i 2,G, . . . , h i C,G) are concatenated to form the final representation vector h′i for wi in the GTN model as h′i=[h i 1,G, h i 2,G, . . . , h i C,G]”).
Regarding claim 8, Pouran in view of Mulligan and Gutierrez teaches all the limitations of claim 7.
Pouran teaches:
wherein aggregating over the lower layer embedding vectors comprises: aggregating over the lower layer embedding vectors of neighboring documents for each respective linking function (Pouran ¶0114: “The event argument extraction network assembles a representation vector R based on the hidden vectors for wa and wt from the GTN model as follows: R=[h′ a ,h′ t,MaxPool(h′ 1 ,h′ 2 , . . . ,h′ N)] (9)”—[wherein the system uses max pooling to aggregate over all the lower layer vectors with respect to each GCN]); and
aggregating over all linking functions in the library of linking functions (Pouran ¶0111: “A ReLU layer may implement a rectified linear activation function, which comprises a piecewise linear function that outputs the input directly if is positive, otherwise, it outputs zero. A rectified linear activation function may be used as a default activation function for many types of neural networks. Using a rectified linear activation function may enable the use of stochastic gradient descent with backpropagation of errors to train deep neural networks. The rectified linear activation function may operate similar to a linear function, but it may enable complex relationships in the data to be learned”; see also Pouran ¶0052: “One or more embodiments of the present disclosure combine different information sources to generate effective document structures for event argument extraction tasks. The event argument extraction network 315 produces document structures based on knowledge from syntax (i.e., dependency trees), discourse (i.e., coreference links), and semantic similarity. Semantic similarity depends on contextualized representation vectors to compute interaction scores between nodes and relies on using external knowledge bases to enrich document structures for event argument extraction. The words in the documents are linked to the entries in one or more external knowledge bases and exploit the entry similarity in knowledge bases to obtain word similarity scores for the structures. In some examples, lexical database (e.g., WordNet) is used as the knowledge base and tools (e.g., word sense disambiguation or WSD) are applied to facilitate word-entry linking. The linked entry or node in WordNet can provide expert knowledge on the meanings of the words (e.g., glossary and hierarchy information). Such expert knowledge complements the contextual information of words and enhances the semantic-based document structures for event argument extraction. In some embodiments, the event argument extraction network 315 uses one or more external knowledge bases for document structures in information extraction tasks”—[(emphasis added) wherein the system uses external sources (e.g., expert external knowledge bases) to link the documents based on dependency trees, discourse, and semantic similarity which all use different functions to find matches, and wherein when the system creates its graph structures, processing them through the GCN, it includes all the previously calculated ReLU outputs])).
Regarding claim 9, Pouran in view of Mulligan and Gutierrez teaches all the limitations of claim 4.
Pouran teaches:
wherein the link scoring model comprises a neural network (Pouran ¶0092–0093: “In some examples, the syntax-based document structure is based on sentence-level event argument extraction where dependency parsing trees of input sentences reveal important context, i.e., via the shortest dependency paths to connect event triggers and arguments, and guide the interaction modeling between words for argument role prediction by LSTM cells. The dependency trees for the sentences in D are used to provide information for the document structures for event argument extraction. In some embodiments, the dependency relations or connections between pairs of words in W are leveraged to compute interaction scores”—[wherein the system uses LSTM cells (i.e., neural network link scoring model) and the dependency relations to compute interaction scores (i.e., compute link scores with a set of second parameters)]).
Regarding claim 13, Pouran in view of Mulligan teaches all the limitations of claim 1.
Pouran in view of Mulligan does not appear to explicitly teach:
accommodating a new document by: applying the expert linking functions to the new document to extract new links between the new document and the documents; and
generating a new representation for the new document based on the representations of the documents that are stored in cache.
However, Gutierrez teaches:
accommodating a new document by: applying a library of linking functions to the new document to extract new links between the new document and the documents (Gutierrez ¶0624: “For instance, the analytics server may retrieve the user's recorded history and any corresponding metadata according to the user's configured settings, and augment or modify it as specified (e.g., with standard metadata about the user's location, device, and other attributes discussed herein). The analytics server may also classify what type of data is being monitored/retrieved. Based on the collected data, the analytics server may also search for additional metadata for a given historical record from third-party sources (e.g., find a URL in a user's browsing history and access that website remotely to gather additional metadata). The analytics server may then interrelate the collected data to other related data within the nodal data structure. As a result, the analytics server may generate/revise relationships between what the user has historically done (e.g., electronic content accessed by the user) and other content within the nodal data structure”—[wherein when new data is found (i.e., accommodating a new document) the system generates/revises (i.e., applies a library of linking functions to the new document) to update the links]); and
generating a new representation for the new document based on the representations of the documents that are stored in cache (Gutierrez ¶0624: “For instance, the analytics server may retrieve the user's recorded history and any corresponding metadata according to the user's configured settings ... The analytics server may then interrelate the collected data to other related data within the nodal data structure. As a result, the analytics server may generate/revise relationships between what the user has historically done (e.g., electronic content accessed by the user) and other content within the nodal data structure”; see also Gutierrez ¶0490: “In some embodiments, the analytics server may utilize a local-first data storage strategy, enabling a user's nodal data structure to be entirely offline, private and controlled by the user … In addition, the extension may store a local cache in IndexedDB of the user's preferences, applications, contexts, and histories of recent browser tabs, which it keeps synchronized with the cloud when either end is updated”—[wherein the system interrelates the new data with the old data (i.e., generating a new representation) based on the old documents and history (i.e., documents stored in cache)]).
The same motivation that was utilized for combining Pouran with Gutierrez, as set forth in claim 4, is equally applicable to claim 13.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NICHOLAS SHINE whose telephone number is (571)272-2512. The examiner can normally be reached M-F, 9a-5p 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, David Yi can be reached on (571) 270-7519. 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.
/N.B.S./Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126