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 .
This action is responsive to the application filed on March 28th, 2023. Claims 1-20 are pending in the case. Claims 1, 8, and 15 are independent claims.
The information disclosure statements (IDS) submitted on 28 March 2023 and 24 July 2023 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1:
Step 1: Claim 1 is directed to A computer-implemented method, therefore it falls under the statuary category of a process.
Step 2A Prong 1: The claim recites, in part:
“generating…node embeddings comprising node pairs of a first schema and a second schema” this limitation is a mathematical concept.
“predicting…a label output for the node pairs” this encompasses the mental prediction of a label output for observed node pairs.
“clustering...the node pairs into a cluster output” this limitation is a mathematical concept.
“determining…that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs” this encompasses the mental determination that an observed label output and an observed cluster output are in a disagreement.
Step 2A Prong 2: The judicial exception is not integrated into a practical application; the remaining limitations of the claim are as follows: “by a first machine learning model executed on a processor”, “by a second machine learning model executed on the processor”, “by the processor” (line XXX of claim 1), “by the processor” (line XXX of claim 1) “in response to displaying the at least one node pair to a subject matter expert to generate a label for the at least one node pair, using, by the processor, the label for the at least one node pair as training data to further train the second machine learning model” the limitations are an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP § 2106.05(f)(2).
Step 2B: The additional elements, taken individually and in combination, do not provide an inventive concept of significantly more than the abstract idea itself for the reasons set forth in step 2A prong 2 above. Therefore, the claim is ineligible.
Regarding claim 2, the rejection of claim 1 is incorporated and further:
Step 2A Prong 1: The claim recites, in part:
“determining that unlabeled node pairs of the node pairs are semantically similar to the at least one node pair” this encompasses the mental determination that an observed unlabeled node pairs of the node pairs is semantically similar to the at least one node pair.
“labeling the unlabeled node pairs having been determined with the label of the at least one node pair” this encompasses the mental labeling of observed unlabeled node pairs.
Step 2A Prong 2: The claim does not recite any additional limitations, thus does not further recite any additional elements that integrates the judicial exception into a practical application or amount to significantly more.
Regarding claim 3, the rejection of claim 1 is incorporated and further:
Step 2A Prong 1: The claim recites, in part:
“generating labeled node pairs by labeling unlabeled node pairs with the label of the at least one node pair, in response to the unlabeled node pairs being semantically similar to the at least one node pair” this encompasses the mental creation of labeled node pairs by labeling observed unlabeled node pairs in response to observing unlabeled node pairs are semantically similar to the at least one node pair.
Step 2A Prong 2: The judicial exception is not integrated into a practical application; the remaining limitations of the claim are as follows: “using the labeled node pairs having the label as further training data to train the second machine learning model” the limitation is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP § 2106.05(f)(2).
Step 2B: The additional elements, taken individually and in combination, do not provide an inventive concept of significantly more than the abstract idea itself for the reasons set forth in step 2A prong 2 above. Therefore, the claim is ineligible.
Regarding claim 4, the rejection of claim 3 is incorporated and further:
Step 2A Prong 1: The claim recites, in part:
“the labeled node pairs are applied at an adaptive rate…, the adaptive rate increasing with each iteration of aligning the first schema and the second schema” this encompasses the mental adaptation of a rate of applying observed labeled node pairs,
Step 2A Prong 2: The judicial exception is not integrated into a practical application; the remaining limitations of the claim are as follows: “as the further training data for training the second machine learning model” the limitation is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP § 2106.05(f)(2).
Step 2B: The additional elements, taken individually and in combination, do not provide an inventive concept of significantly more than the abstract idea itself for the reasons set forth in step 2A prong 2 above. Therefore, the claim is ineligible.
Regarding claim 5, the rejection of claim 1 is incorporated and further:
Step 2A Prong 1: The claim recites, in part:
“comparing a model similarity score associated with the label output to a clustering similarity score associated with the cluster output for the at least one node pair” this encompasses the mental comparison of observed similarity scores.
“determining that a difference in the model similarity score and the clustering similarity score is greater than a threshold” this encompasses the mental determination that a difference in observed similarity scores is greater than a threshold. Further, this limitation is a mathematical concept.
Step 2A Prong 2: The claim does not recite any additional limitations, thus does not further recite any additional elements that integrates the judicial exception into a practical application or amount to significantly more.
Regarding claim 6, the rejection of claim 1 is incorporated and further:
Step 2A Prong 1: a continuation of the abstract idea identified in the parent claim.
Step 2A Prong 2: The judicial exception is not integrated into a practical application; the remaining limitations of the claim are as follows: “the first machine learning model comprises a relational graph convolution network” the limitation is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(h).
Step 2B: The additional elements, taken individually and in combination, do not provide an inventive concept of significantly more than the abstract idea itself for the reasons set forth in step 2A prong 2 above. Therefore, the claim is ineligible.
Regarding claim 7, the rejection of claim 1 is incorporated and further:
Step 2A Prong 1: a continuation of the abstract idea identified in the parent claim.
Step 2A Prong 2: The judicial exception is not integrated into a practical application; the remaining limitations of the claim are as follows: “the second machine learning model comprises a classifier” the limitation is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(h).
Step 2B: The additional elements, taken individually and in combination, do not provide an inventive concept of significantly more than the abstract idea itself for the reasons set forth in step 2A prong 2 above. Therefore, the claim is ineligible.
Regarding claim 8:
Step 1: Claim 8 is directed to A system, therefore it falls under the statuary category of a machine.
Step 2A Prong 1: The claim recites, in part:
“generating…node embeddings comprising node pairs of a first schema and a second schema” this limitation is a mathematical concept.
“predicting…a label output for the node pairs” this encompasses the mental prediction of a label output for observed node pairs.
“clustering the node pairs into a cluster output” this limitation is a mathematical concept.
“determining that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs” this encompasses the mental determination that an observed label output and an observed cluster output are in a disagreement.
Step 2A Prong 2: The judicial exception is not integrated into a practical application; the remaining limitations of the claim are as follows: “a memory having computer readable instructions” the limitation is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(h). “a computer for executing the computer readable instructions, the computer readable instructions controlling the computer to perform operations comprising”, “by a first machine learning model”, “by a second machine learning model”, “in response to displaying the at least one node pair to a subject matter expert to generate a label for the at least one node pair, using the label for the at least one node pair as training data to further train the second machine learning model” the limitations are an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP § 2106.05(f)(2).
Regarding claims 9-14:
The rejection of claim 8 is further incorporated, the rejection of claims 2-7 are applicable to claims 9-14, respectively.
Regarding claim 15:
Step 1: Claim 15 is directed to A computer program product, therefore it falls under the statuary category of a manufacture.
Step 2A Prong 1: The claim recites, in part:
“generating…node embeddings comprising node pairs of a first schema and a second schema” this limitation is a mathematical concept.
“predicting…a label output for the node pairs” this encompasses the mental prediction of a label output for observed node pairs.
“clustering the node pairs into a cluster output” this limitation is a mathematical concept.
“determining that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs” this encompasses the mental determination that an observed label output and an observed cluster output are in a disagreement.
Step 2A Prong 2: The judicial exception is not integrated into a practical application; the remaining limitations of the claim are as follows: “a computer readable storage medium having program instructions embodied therewith” the limitation is an additional element that generally links the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(h). “the program instructions executable by a computer to cause the computer to perform operations comprising”, “by a first machine learning model”, “by a second machine learning model”, “in response to displaying the at least one node pair to a subject matter expert to generate a label for the at least one node pair, using the label for the at least one node pair as training data to further train the second machine learning model” the limitations are an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP § 2106.05(f)(2).
Regarding claims 16-20:
The rejection of claim 15 is further incorporated, the rejection of claims 2-6 are applicable to claims 16-20, respectively.
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.
Claims 1, 5-8, 12-15 and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Hao et al. ("MEDTO: Medical Data to Ontology Matching Using Hybrid Graph Neural Networks", Hao et al., 14 August 2021), as cited in the IDS, hereinafter Hao in view of McCallum et al. ("Efficient clustering of high-dimensional data sets with application to reference matching", McCallum et al., 01 August 2000) hereinafter McCallum in further view of Caramalau et al. ("Sequential Graph Convolutional Network for Active Learning", Caramalau et al., 1 Apr 2021) hereinafter Caramalau.
Regarding claim 1:
Hao teaches A computer-implemented method comprising:
generating, by a first machine learning model executed on a processor, node embeddings comprising node pairs of a first schema and a second schema (Hao, page 4, col 1, ¶2 “The graph encoder uses the graph structure to propagate and aggregate information across nodes and learn embeddings that encode local structural information. A graph decoder is often used to compute similarity scores for all node pairs for downstream tasks on node, edge, or graph level.”);
predicting, by a second machine learning model executed on the processor, a label output for the node pairs (Hao, page 7, col 1, section 4.4, ¶1 “Then, the matching module 𝑀(·) takes pairs of concept embeddings from O1 and O2 and outputs the prediction score.”);
in response to displaying the at least one node pair to a subject matter expert to generate a label for the at least one node pair (Hao, page 8, col 1, section 5.4, ¶1 “We evaluate Medto on both MIMIC-III and MDX. For MIMIC-III, our domain experts identified 15 matching concepts in SNOMED, among 21 tables.”)
Hao does not teach "clustering, by the processor, the node pairs into a cluster output;
determining, by the processor, that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs"
However, McCallum teaches clustering, by the processor, the node pairs into a cluster output (McCallum, page 5, col 1, ¶2 “Initialize each element to be a cluster of size one, compute the distances between all pairs of clusters, sort the distances from smallest to largest, and then repeatedly merge the two clusters which are closest together until one is left with the desired number of clusters.” Here, the desired number of clusters can be considered the cluster output);
determining, by the processor, that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs (McCallum, page 7, col 2, ¶4 “Thus, the error rate is the fraction of pairs of citations that are correctly in the same cluster (if they reference the same paper) or in different clusters (if they reference different papers).” Here, the error rate can be considered the disagreement, and the correct citations can be considered the label output and the clusters can be considered the cluster output);
Hao and McCallum are analogous art because both references concern methods for reference matching. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao’s ontology matching to incorporate the clustering canopies taught by McCallum. The motivation for doing so would have been to provide a principled approach to large dataset problems as stated in McCallum, page 10, col 1, section V, ¶2 “Given large data sets with hundreds of thousands or millions of entries, computing all pairwise similarities between objects is often intractable, and more efficient methods are called for. Also, increasingly, people are trying to fit complex models such as mixture distributions or HMMs to these large data sets. Computing global models where all observations can affect all parameters is again intractable, and methods for grouping observations (similarly to the grouping of objects above) are needed. Canopies provide a principled approach to all these problems.”.
Hao in view of McCallum does not teach "using, by the processor, the label for the at least one node pair as training data to further train the second machine learning model"
However, Caramalau teaches using, by the processor, the label for the at least one node pair as training data to further train the second machine learning model (Caramalau, page 1, figure 1
PNG
media_image1.png
181
342
media_image1.png
Greyscale
and further Caramalau, page 2, col 2, ¶1 “Traversing from Phase I to Phase V as shown in Figure 1 completes a cycle. In the next iteration, we flip the label of annotated examples from unlabelled to labelled and re-train the whole framework.” here, in response to the sample being labelled by the oracle, the machine learning model is trained).
Hao in view of McCallum and Caramalau are analogous art because both references concern methods for graph convolution networks. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao/McCallum’s graph convolution network to incorporate the oracle based retraining taught by Caramalau. The motivation for doing so would have been to incorporate the clear advantages of their GCN, as stated in Caramalau, page 2, col 2, ¶1 “Our method has a clear advantage due to the aforementioned strengths of the GCN which is demonstrated by both the quantitative and qualitative experiments (see Section 4). Traversing from Phase I to Phase V as shown in Figure 1 completes a cycle. In the next iteration, we flip the label of annotated examples from unlabelled to labelled and re-train the whole framework.”
Regarding claim 5:
Hao in view of McCallum in further view of Caramalau teaches The computer-implemented method of claim 1, wherein determining that the label output and the cluster output are in the disagreement for the at least one node pair of the node pairs comprises: comparing a model similarity score associated with the label output to a clustering similarity score associated with the cluster output for the at least one node pair, and determining that a difference in the model similarity score and the clustering similarity score is greater than a threshold (McCallum, page 7, col 2, ¶4 “Thus, the error rate is the fraction of pairs of citations that are correctly in the same cluster (if they reference the same paper) or in different clusters (if they reference different papers).” Here, the error rate can be considered the disagreement, and the correct citations can be considered the label output and the clusters can be considered the cluster output further McCallum, page 8, col 2, ¶2 “In performing a canopy clustering, two sets of parameters need to be chosen: the values for the two canopy thresholds, and the number of final clusters”).
It would have been obvious to combine the teachings of Hao in view of McCallum and Caramalau teaches for the reasons set forth in connection with claim 1 above.
Regarding claim 6:
Hao in view of McCallum in further view of Caramalau teaches The computer-implemented method of claim 1, wherein the first machine learning model comprises a relational graph convolution network (Hao, page 6, col 1, section 4.3, ¶1 “To capture the non-hierarchical structure in an ontology, conventional GNNs such as R-GCN [31] can be applied, as it models multirelational graphs”).
Regarding claim 7:
Hao in view of McCallum in further view of Caramalau teaches The computer-implemented method of claim 1, wherein the second machine learning model comprises a classifier (Hao, page 7, col 1, section 4.4, ¶1 “Then, the matching module 𝑀(·) takes pairs of concept embeddings from O1 and O2 and outputs the prediction score. We use the straightforward multi-layer perceptron (MLP) with one hidden layer…” in light of ¶55 of the specification, the MLP can be considered a classifier “The machine learning model234 can be a classifier such as a multilayer perceptron (MLP).”).
Regarding claim 8:
Hao teaches A system comprising:
a memory having computer readable instructions; and a computer for executing the computer readable instructions, the computer readable instructions controlling the computer to perform operations comprising (Hao, page 8, col 1, section 5.4, ¶1 “We evaluate Medto on both MIMIC-III and MDX. For MIMIC-III, our domain experts identified 15 matching concepts in SNOMED, among 21 tables. For MDX, 19 out of 59 tables have their matches identified in SNOMED as well” here, evaluation on the MIMIC-III dataset inherently discloses execution on a computing device):
generating, by a first machine learning model executed on a processor, node embeddings comprising node pairs of a first schema and a second schema (Hao, page 4, col 1, ¶2 “The graph encoder uses the graph structure to propagate and aggregate information across nodes and learn embeddings that encode local structural information. A graph decoder is often used to compute similarity scores for all node pairs for downstream tasks on node, edge, or graph level.”);
predicting, by a second machine learning model executed on the processor, a label output for the node pairs (Hao, page 7, col 1, section 4.4, ¶1 “Then, the matching module 𝑀(·) takes pairs of concept embeddings from O1 and O2 and outputs the prediction score.”);
in response to displaying the at least one node pair to a subject matter expert to generate a label for the at least one node pair (Hao, page 8, col 1, section 5.4, ¶1 “We evaluate Medto on both MIMIC-III and MDX. For MIMIC-III, our domain experts identified 15 matching concepts in SNOMED, among 21 tables.”)
Hao does not teach "clustering the node pairs into a cluster output;
determining that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs"
However, McCallum teaches clustering the node pairs into a cluster output (McCallum, page 5, col 1, ¶2 “Initialize each element to be a cluster of size one, compute the distances between all pairs of clusters, sort the distances from smallest to largest, and then repeatedly merge the two clusters which are closest together until one is left with the desired number of clusters.” Here, the desired number of clusters can be considered the cluster output);
determining that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs (McCallum, page 7, col 2, ¶4 “Thus, the error rate is the fraction of pairs of citations that are correctly in the same cluster (if they reference the same paper) or in different clusters (if they reference different papers).” Here, the error rate can be considered the disagreement, and the correct citations can be considered the label output and the clusters can be considered the cluster output);
Hao and McCallum are analogous art because both references concern methods for reference matching. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao’s ontology matching to incorporate the clustering canopies taught by McCallum. The motivation for doing so would have been to provide a principled approach to large dataset problems as stated in McCallum, page 10, col 1, section V, ¶2 “Given large data sets with hundreds of thousands or millions of entries, computing all pairwise similarities between objects is often intractable, and more efficient methods are called for. Also, increasingly, people are trying to fit complex models such as mixture distributions or HMMs to these large data sets. Computing global models where all observations can affect all parameters is again intractable, and methods for grouping observations (similarly to the grouping of objects above) are needed. Canopies provide a principled approach to all these problems.”.
Hao in view of McCallum does not teach "using the label for the at least one node pair as training data to further train the second machine learning model"
However, Caramalau teaches using the label for the at least one node pair as training data to further train the second machine learning model (Caramalau, page 1, figure 1
PNG
media_image1.png
181
342
media_image1.png
Greyscale
and further Caramalau, page 2, col 2, ¶1 “Traversing from Phase I to Phase V as shown in Figure 1 completes a cycle. In the next iteration, we flip the label of annotated examples from unlabelled to labelled and re-train the whole framework.” here, in response to the sample being labelled by the oracle, the machine learning model is trained).
Hao in view of McCallum and Caramalau are analogous art because both references concern methods for graph convolution networks. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao/McCallum’s graph convolution network to incorporate the oracle based retraining taught by Caramalau. The motivation for doing so would have been to incorporate the clear advantages of their GCN, as stated in Caramalau, page 2, col 2, ¶1 “Our method has a clear advantage due to the aforementioned strengths of the GCN which is demonstrated by both the quantitative and qualitative experiments (see Section 4). Traversing from Phase I to Phase V as shown in Figure 1 completes a cycle. In the next iteration, we flip the label of annotated examples from unlabelled to labelled and re-train the whole framework.”
Regarding claims 12-14:
Claims 12-14 are rejected under the same rationale as claims 5-7, respectively.
Regarding claim 15:
Hao teaches A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform operations comprising:
generating, by a first machine learning model executed on a processor, node embeddings comprising node pairs of a first schema and a second schema (Hao, page 4, col 1, ¶2 “The graph encoder uses the graph structure to propagate and aggregate information across nodes and learn embeddings that encode local structural information. A graph decoder is often used to compute similarity scores for all node pairs for downstream tasks on node, edge, or graph level.”);
predicting, by a second machine learning model executed on the processor, a label output for the node pairs (Hao, page 7, col 1, section 4.4, ¶1 “Then, the matching module 𝑀(·) takes pairs of concept embeddings from O1 and O2 and outputs the prediction score.”);
in response to displaying the at least one node pair to a subject matter expert to generate a label for the at least one node pair (Hao, page 8, col 1, section 5.4, ¶1 “We evaluate Medto on both MIMIC-III and MDX. For MIMIC-III, our domain experts identified 15 matching concepts in SNOMED, among 21 tables.”)
Hao does not teach "clustering the node pairs into a cluster output;
determining that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs"
However, McCallum teaches clustering the node pairs into a cluster output (McCallum, page 5, col 1, ¶2 “Initialize each element to be a cluster of size one, compute the distances between all pairs of clusters, sort the distances from smallest to largest, and then repeatedly merge the two clusters which are closest together until one is left with the desired number of clusters.” Here, the desired number of clusters can be considered the cluster output);
determining that the label output and the cluster output are in a disagreement for at least one node pair of the node pairs (McCallum, page 7, col 2, ¶4 “Thus, the error rate is the fraction of pairs of citations that are correctly in the same cluster (if they reference the same paper) or in different clusters (if they reference different papers).” Here, the error rate can be considered the disagreement, and the correct citations can be considered the label output and the clusters can be considered the cluster output);
Hao and McCallum are analogous art because both references concern methods for reference matching. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao’s ontology matching to incorporate the clustering canopies taught by McCallum. The motivation for doing so would have been to provide a principled approach to large dataset problems as stated in McCallum, page 10, col 1, section V, ¶2 “Given large data sets with hundreds of thousands or millions of entries, computing all pairwise similarities between objects is often intractable, and more efficient methods are called for. Also, increasingly, people are trying to fit complex models such as mixture distributions or HMMs to these large data sets. Computing global models where all observations can affect all parameters is again intractable, and methods for grouping observations (similarly to the grouping of objects above) are needed. Canopies provide a principled approach to all these problems.”.
Hao in view of McCallum does not teach "using the label for the at least one node pair as training data to further train the second machine learning model"
However, Caramalau teaches using the label for the at least one node pair as training data to further train the second machine learning model (Caramalau, page 1, figure 1
PNG
media_image1.png
181
342
media_image1.png
Greyscale
and further Caramalau, page 2, col 2, ¶1 “Traversing from Phase I to Phase V as shown in Figure 1 completes a cycle. In the next iteration, we flip the label of annotated examples from unlabelled to labelled and re-train the whole framework.” here, in response to the sample being labelled by the oracle, the machine learning model is trained).
Hao in view of McCallum and Caramalau are analogous art because both references concern methods for graph convolution networks. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao/McCallum’s graph convolution network to incorporate the oracle based retraining taught by Caramalau. The motivation for doing so would have been to incorporate the clear advantages of their GCN, as stated in Caramalau, page 2, col 2, ¶1 “Our method has a clear advantage due to the aforementioned strengths of the GCN which is demonstrated by both the quantitative and qualitative experiments (see Section 4). Traversing from Phase I to Phase V as shown in Figure 1 completes a cycle. In the next iteration, we flip the label of annotated examples from unlabelled to labelled and re-train the whole framework.”
Regarding claims 19-20:
Claims 19-20 are rejected under the same rationale as claims 5-6, respectively.
Claims 2-3, 9-10 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Hao in view of McCallum in view of Caramalau in further view of Iscen et al. ("Label Propagation for Deep Semi-supervised Learning", Iscen et al., 9 Apr 2019) hereinafter Iscen.
Regarding claim 2:
Hao in view of McCallum in further view of Caramalau teaches The computer-implemented method of claim 1,
Hao in view of McCallum in further view of Caramalau does not teach "further comprising determining that unlabeled node pairs of the node pairs are semantically similar to the at least one node pair; and
labeling the unlabeled node pairs having been determined with the label of the at least one node pair"
However, Iscen teaches further comprising determining that unlabeled node pairs of the node pairs are semantically similar to the at least one node pair (Iscen, page 7, col 2, ¶3 “Pairwise similarities for the graph are computed with the publicly available FAISS library [21]”); and
labeling the unlabeled node pairs having been determined with the label of the at least one node pair (Iscen, page 7, col 2, ¶3 “Class weights ζj are normalized over c classes such that the average class weight is 1. Pseudo-label predictions, ωi , and ζj are updated after each epoch.”).
Hao in view of McCallum in further view of Caramalau and Iscen are analogous art because both references concern methods for graph based deep learning. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao/McCallum/Caramalau’s ontology matching system to incorporate the label propagation taught by Iscen. The motivation for doing so would have been to outperform other semi-supervised methods as stated in Iscen, page 2, col 1, ¶3 “We experimentally show on standard datasets that the proposed method outperforms other semi-supervised approaches. The less labeled data is available, the more pronounced the advantage of the proposed approach is.”.
Regarding claim 3:
Hao in view of McCallum in further view of Caramalau teaches The computer-implemented method of claim 1,
Hao in view of McCallum in further view of Caramalau does not teach “further comprising generating labeled node pairs by labeling unlabeled node pairs with the label of the at least one node pair, in response to the unlabeled node pairs being semantically similar to the at least one node pair; and
using the labeled node pairs having the label as further training data to train the second machine learning model”
However, Iscen teaches further comprising generating labeled node pairs by labeling unlabeled node pairs with the label of the at least one node pair, in response to the unlabeled node pairs being semantically similar to the at least one node pair (Iscen, page 7, col 2, ¶3 “Pairwise similarities for the graph are computed with the publicly available FAISS library [21]”); and
using the labeled node pairs having the label as further training data to train the second machine learning model (Iscen, page 2, col1 , ¶2 “In this paper, we use efficient transductive label propagation [43] to infer pseudo-labels for unlabeled data, which are used to train the classifier.”).
Hao in view of McCallum in further view of Caramalau and Iscen are analogous art because both references concern methods for graph based deep learning. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao/McCallum/Caramalau’s ontology matching system to incorporate the label propagation taught by Iscen. The motivation for doing so would have been to outperform other semi-supervised methods as stated in Iscen, page 2, col 1, ¶3 “We experimentally show on standard datasets that the proposed method outperforms other semi-supervised approaches. The less labeled data is available, the more pronounced the advantage of the proposed approach is.”.
Regarding claims 9-10 and 16-17:
Claims 9-10 and 16-17 are rejected under the same rationale as claims 2-3, respectively.
Claims 4, 11 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Hao in view of McCallum in view of Caramalau in view of Iscen in further view of Zhang et al. ("FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling", Zhang et al., 29 Jan 2022) hereinafter Zhang.
Regarding claim 4:
Hao in view of McCallum in view of Caramalau in further view of Iscen teaches The computer-implemented method of claim 3, wherein
Hao in view of McCallum in view of Caramalau in further view of Iscen does not teach "the labeled node pairs are applied at an adaptive rate as the further training data for training the second machine learning model, the adaptive rate increasing with each iteration of aligning the first schema and the second schema"
However, Zhang teaches the labeled node pairs are applied at an adaptive rate as the further training data for training the second machine learning model, the adaptive rate increasing with each iteration of aligning the first schema and the second schema (Zhang, page 6, section 4, ¶3 “For all datasets, we use an initial learning rate of 0.03 with a cosine learning rate decay schedule [31] as η = η0 cos( 7πk 16K ), where η0 is the initial learning rate, k is the current training step and K is the total training step that is set to 2 20. We also perform an exponential moving average with the momentum of 0.999. The batch size of labeled data is 64 except for ImageNet. µ is set to be 1 for Pseudo-Label and 7 for UDA, FixMat”).
Hao in view of McCallum in view of Caramalau in further view of Iscen and Zhang are analogous art because both references concern methods for reference matching. Accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Hao/McCallum/Caramalau/Iscen’s ontology matching method to incorporate the iterative training taught by Zhang. The motivation for doing so would have been to achieve strong performances when the labeled data are extremely limited or when the task is challenging as stated in Zhang, page 1, abstract “FlexMatch achieves state-of-the-art performance on a variety of SSL benchmarks, with especially strong performances when the labeled data are extremely limited or when the task is challenging.”.
Regarding claims 11 and 18:
Claims 11 and 18 are rejected under the same rationale as claim 4.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Liu et al. ("GIANT: Scalable Creation of a Web-scale Ontology", Liu et al., 11 June 2020) teaches a mechanism to construct a user-centered, web-scale, structured ontology, containing a large number of natural language phrases conforming to user attentions at various granularities, mined from a vast volume of web documents and search click graphs. Various types of edges are also constructed to maintain a hierarchy in the ontology.
Zhu et al. ("Learning from Labeled and Unlabeled Data with Label Propagation", Zhu et al., June 2002) teaches an iterative method of propagating labeled data to label a dataset.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JACOB Z SUSSMAN MOSS whose telephone number is (571) 272-1579. The examiner can normally be reached Monday - Friday, 9 a.m. - 5 p.m. 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, Kakali Chaki can be reached on (571) 272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/J.S.M./Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122