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 .
Applicant has canceled claims 11-14 and added new claims 15-34 on the preliminary claims filed on 11/20/2025.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 28 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 28 recites “wherein the uncertainty value has limits from 0 to 1, and it is less than 0.5, or less than 0.35, or less than 0.25, or around 0.05 and 0.15” which is indefinite.
The term “around” is a relative term that renders the scope of the claim unclear because the claim does not provide a boundary for determining the values encompassed by “around 0.05 and 0.15” or “around 0.1.” For example, it is unclear whether 0.095 is considered “around 0.1,” and it is further unclear how far above or below 0.1 a value may be and still satisfy the limitation.
The Specification at page 13 states that the uncertainty value “has limits from 0 to 1, and it is preferably less than 0.5, more preferably less than 0.35, even more preferably less than 0.25, most preferably around 0.05 and 0.15, preferably around 0.1.” The Specification further provides an example in which an uncertainty value of 0.05 corresponds to a query result having an error of equal to or less than 5%. However, these disclosures do not establish an objective numerical boundary for the term “around.” Rather, the values 0.05, 0.15, and 0.1 are presented as preferred values/examples within the disclosed uncertainty range without specifying the amount of permissible deviation from those values.
For purposes of examination, the Examiner interprets “around 0.05 and 0.15” and “around 0.1” broadly as encompassing uncertainty values reasonably near the recited values and within the disclosed uncertainty range.
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 15-34 are rejected under 35 U.S.C. 101 because of the following reasons:
Claims 15, 34:
At Step 1:
The claim is directed to "a methods" and “a non-transitory computer readable medium” and thus directed to a statutory category.
At Step 2A, Prong One:
The claim recites the following limitations directed to an abstract idea:
-“reading a query
-“choosing an optimal way of processing the query by utilizing an uncertainty value to provide a search strategy” recites a mental process because human mind can select (choose) and an optimal way of processing a query based on the uncertainty value to provide a search strategy by evaluation and judgement of data.
-“based on the search strategy to traverse at least a part of a data storage and utilize at least a part of an embedding storage” recites a mental process because human mind can traverse a data storage and utilize an embedding storage based on the search strategy by evaluation/analysis of data.
At Step 2A, Prong Two:
The claim recites the following additional elements:
- "a computer implemented method for handling queries, the method comprising the steps of", “a non-transitory computer-readable medium storing instruction that, when executed by one or more processors, cause the one or more processors to carry out the method of handling queries comprising the steps of” which are all a high-level recitation of a generic computer components and represent mere instructions to apply the judicial exception on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
-"by using a query parser", “a hybrid query optimizer to provide a search strategy”, “based on the search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse
-“wherein the data storage comprises base data stored in a graph format and the embedding storage comprises supplementary generated data for at least a part of the base data, wherein the supplementary generated data have been generated using Machine Learning” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
-“returning a query result based on the uncertainty value and the search strategy” is insignificant extra-solution activity as outputting data. See MPEP 2106.05(g).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-"by using a query parser", “a hybrid query optimizer to provide a search strategy”, “based on the search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse
-“wherein the data storage comprises base data stored in a graph format and the embedding storage comprises supplementary generated data for at least a part of the base data, wherein the supplementary generated data have been generated using Machine Learning” are generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
-“returning a query result based on the uncertainty value and the search strategy” is well-known, routine and conventional activities (WURC), see MPEP 2106.05(d)(II) by at least "i. Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, buySAFE, Inc. V. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network)" and/or "iv. Presenting offers and gathering statistics, OIP Techs., 788 F.3d at 1362-63, 115 USPQ2d at 1092-9".
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 16:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the base data are stored in a generic data model supporting graph” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h)
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the base data are stored in a generic data model supporting graph” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h)
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 17:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the query result is stored as new data in incomplete graphs of the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type. (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the query result is stored as new data in incomplete graphs of the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type. (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 18:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the uncertainty value comprises a global uncertainty value set at a global level of the query, and/or a local uncertainty value, set at a local level of the query” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the uncertainty value comprises a global uncertainty value set at a global level of the query, and/or a local uncertainty value, set at a local level of the query” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type(technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 19:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the supplementary data in the embedding storage are compressed vector representations of all the nodes of the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the supplementary data in the embedding storage are compressed vector representations of all the nodes of the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h)
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 20:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the embedding storage is continuously supplemented with additional supplementary data based on the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the embedding storage is continuously supplemented with additional supplementary data based on the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 21:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the query has any graph query language format” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the query has any graph query language format” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 22:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the uncertainty value determines the amount of processes performed by the computer implemented method on the data storage and on the embedding storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the uncertainty value determines the amount of processes performed by the computer implemented method on the data storage and on the embedding storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 23:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein a cost model is included in the search strategy of the hybrid query optimizer, which reflects the time needed for the method to provide a query under a given uncertainty value” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein a cost model is included in the search strategy of the hybrid query optimizer, which reflects the time needed for the method to provide a query under a given uncertainty value” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 24:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the hybrid query optimizer enumerates candidate hybrid plans, which find the optimal cost for a given uncertainty value for a specific query” ” is apply it limitation of using the computer as a tool, i.e ML-based query executor, to perform the judicial exception (2106.05(f)).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the hybrid query optimizer enumerates candidate hybrid plans, which find the optimal cost for a given uncertainty value for a specific query” ” is apply it limitation of using the computer as a tool, i.e ML-based query executor, to perform the judicial exception (2106.05(f)).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 25:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the embedding storage comprises supplementary generated data based on incomplete graphs of the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the embedding storage comprises supplementary generated data based on incomplete graphs of the data storage” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 26:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the query comprises a plurality of operators that execute on a graph” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the query comprises a plurality of operators that execute on a graph” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 27:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the plurality of operators that execute on a graph are node distance approximates” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the plurality of operators that execute on a graph are node distance approximates” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 28:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the uncertainty value has limits from 0 to 1, and it is less than 0.5, or less than 0.35, or less than 0.25, or around 0.05 and 0.15,or around 0.1” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the uncertainty value has limits from 0 to 1, and it is less than 0.5, or less than 0.35, or less than 0.25, or around 0.05 and 0.15,or around 0.1” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 29:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the generic data model supporting graph is a property graph data model or a relational model” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the generic data model supporting graph is a property graph data model or a relational model” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 30:
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“wherein the uncertainty value is provided by a user or by the hybrid query optimizer” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“wherein the uncertainty value is provided by a user or by the hybrid query optimizer” is generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 31:
At Step 2A, Prong One:
The claim recites the following limitations directed to an abstract idea:
-“comprising the step of choosing one or more search strategies for processing the query wherein each search strategy is associated with an uncertainty value” recites a mental process because human mind can select one or more search strategies for the query which are associated with uncertainty value by evaluation and judgement of data.
At Step 2A, Prong Two:
The claim recites the following additional elements:
-“by utilizing a hybrid query optimizer” is apply it limitation of using the computer as a tool, i.e ML-based query executor, to perform the judicial exception (2106.05(f)).
-“and returning a query result and an uncertainty value for each search strategy” is insignificant extra-solution activity as outputting data. See MPEP 2106.05(g).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“by utilizing a hybrid query optimizer” is apply it limitation of using the computer as a tool, i.e ML-based query executor, to perform the judicial exception, See MPEP (2106.05(f)).
-“and returning a query result and an uncertainty value for each search strategy” is well-known, routine and conventional activities (WURC), See MPEP 2106.05(d)(II) by at least "i. Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, buySAFE, Inc. V. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network)" and/or "iv. Presenting offers and gathering statistics, OIP Techs., 788 F.3d at 1362-63, 115 USPQ2d at 1092-9".
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 32:
At Step 1:
The claim is directed to “a computer implemented method for handling queries” and thus directed to a statutory category.
At Step 2A, Prong One:
The claim recites the following limitations directed to an abstract idea:
-“reading a query
-“choosing an optimal way of processing the query by utilizing an uncertainty value to provide a search strategy, wherein each search strategy is associated with an uncertainty value” recites a mental process because human mind can select (choose) and an optimal way of processing a query based on the uncertainty value to provide a search strategy by evaluation and judgement of data.
-“based on the search strategy to traverse at least a part of a data storage and utilize at least a part of an embedding storage” recites a mental process because human mind can traverse a data storage and utilize an embedding storage based on the search strategy by evaluation/analysis of data.
At Step 2A, Prong Two:
The claim recites the following additional elements:
- "a computer implemented method for handling queries, the method comprising the steps of” which are all a high-level recitation of a generic computer components and represent mere instructions to apply the judicial exception on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
-"by using a query parser", “a hybrid query optimizer to provide a search strategy”, “based on the search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse
-“wherein the data storage comprises base data stored in a graph format and the embedding storage comprises supplementary generated data for at least a part of the base data, wherein the supplementary generated data have been generated using Machine Learning” are generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
-“returning a query result and an uncertainty value for each search strategy” is insignificant extra-solution activity as outputting data. See MPEP 2106.05(g).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-"by using a query parser", “a hybrid query optimizer to provide a search strategy”, “based on the search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse
-“wherein the data storage comprises base data stored in a graph format and the embedding storage comprises supplementary generated data for at least a part of the base data, wherein the supplementary generated data have been generated using Machine Learning” are generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
-“returning a query result based on the uncertainty value and the search strategy” is well-known, routine and conventional activities (WURC), See MPEP 2106.05(d)(II) by at least "i. Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, buySAFE, Inc. V. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network)" and/or "iv. Presenting offers and gathering statistics, OIP Techs., 788 F.3d at 1362-63, 115 USPQ2d at 1092-9".
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
Claim 33:
At Step 1:
The claim is directed to "a query handling system" and thus directed to a statutory category.
At Step 2A, Prong One:
The claim recites the following limitations directed to an abstract idea:
At Step 2A, Prong One:
The claim recites the following limitations directed to an abstract idea:
-“choosing an optimal way of processing the query by utilizing an uncertainty value to provide a search strategy” recites a mental process because human mind can select (choose) and an optimal way of processing a query based on the uncertainty value to provide a search strategy by evaluation and judgement of data.
-“based on the search strategy to traverse at least a part of a data storage and utilize at least a part of an embedding storage” recites a mental process because human mind can traverse a data storage and utilize an embedding storage based on the search strategy by evaluation/analysis of data.
At Step 2A, Prong Two:
The claim recites the following additional elements:
- "a query handling system”, “a server unit for storing data”, “a processing unit”, which are all a high-level recitation of a generic computer components and represent mere instructions to apply the judicial exception on a computer as in MPEP 2106.05(f), which does not provide integration into a practical application.
-“obtain a query” is insignificant extra-solution activity as data gathering. See MPEP 2106.05(g).
-“a hybrid query optimizer to provide a search strategy”, “based on the search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse
-“wherein the data storage comprises base data stored in a graph format and the embedding storage comprises supplementary generated data for at least a part of the base data, wherein the supplementary generated data have been generated using Machine Learning” are generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
-“returning a query result based on the uncertainty value and the search strategy” is insignificant extra-solution activity as outputting data. See MPEP 2106.05(g).
Viewing the additional limitations together and the claim as a whole, nothing provides integration into a practical application.
At Step 2B:
The conclusions for the mere implementation using a computer are carried over and do not provide significantly more.
-“obtain a query” is well-known, routine and conventional activities (WURC) See MPEP 2106.05(d)(II) by at least "i. Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, buySAFE, Inc. V. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network)".
-"by using a query parser", “a hybrid query optimizer to provide a search strategy”, “based on the search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse limitation of using the computer as a tool, i.e ML-based query executor, to perform the judicial exception (2106.05(f)).
-“wherein the data storage comprises base data stored in a graph format and the embedding storage comprises supplementary generated data for at least a part of the base data, wherein the supplementary generated data have been generated using Machine Learning” are generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type (technical field of ML-based query executor/selector environment). See MPEP § 2106.05(h).
-“returning a query result based on the uncertainty value and the search strategy” is well-known, routine and conventional activities (WURC) See MPEP 2106.05(d)(II) by at least "i. Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, buySAFE, Inc. V. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network)" and/or "iv. Presenting offers and gathering statistics, OIP Techs., 788 F.3d at 1362-63, 115 USPQ2d at 1092-9".
Accordingly, at step 2B, these additional elements, both individually and in
combination, do not amount to significantly more than the judicial exception. See MPEP
§ 2106.05. Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
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.
Claim(s) 15, 16, 19, 21, 23, 24, 26, 29, 30, 31, 32, 33, 34 is/are rejected under 35 U.S.C. 103 as being unpatentable over, NPL “Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection”, by Xiang Yu filed on September 1, 2022, pages 3924-3936 and in view of “Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases” by Hongyu Ren filed on March 26, 2023 pages 1-65.
With respect to claim 15, Yu teaches a computer implemented method for handling queries, the method comprising the steps of ([page 3926, col. 1, element: 2.2 Hybrid00 Framework; “we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models. As shown in Figure 2, the entire framework consists of three key modules”]; selecting queries to process (handling queries)):
reading a query ([page 3926, col. 1, element 2.2, “we leverage a learning-based uncertainty model to select the best plan from candidate plans. Finally, we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], receiving the query and selecting the best plan to execute the query (reading the query));
choosing an optimal way of processing the query by utilizing an uncertainty value and a hybrid query optimizer to provide a search strategy ([page 3925, col. 1, last paragraph “We propose an adaptive hybrid optimizer that combines the cost based and learning-based optimizer through the hint functionality”], [page 3926, col. 1, element 2.2, “Hybrid00 Framework”; “the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty]; [page 3930, col. 1, element 4.4, Uncertainty-based Optimal Plan Selection; “We first filter out the learning-based plans with high uncertainty to obtain the remaining 𝐻 accurately estimated plans, each of which is denoted”], [page 3930, col. 2, “then use the mapping to keep (drop) low (high) uncertainty queries”], selecting the optimal plan from the candidate plans (optimal way to provide a search strategy) with low uncertainty (uncertainty value) and the candidate plans are generated and evaluated using a hybrid query optimizer);
based on the search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse at least a part of a data (page 3926, col. 1, element 2.2, “As it is expensive to fully traverse the leading hint tree, we propose a learning-based leading hint selection method 𝑖.𝑒., MCTS, to select leading hints. Then, we use the hint functionality in DBMS to generate complete plans as candidate plans”], [page 3926, col. 1, element 2.2, “query. Then we first select leading hints from the leading hint tree by MCTS which is a typical tree search algorithm. We will discuss how to use MCTS to select leading hints later. For each selected leading hint, 𝑖.𝑒., the left-deep join orders of the prefixes, we leverage the cost-based optimizer in DBMS to produce a complete query plan. For each hint, we generate a candidate plan”], [page 3930, col. 2, element 4.5, “As shown in Figure 2, HybridQO collects all training data with a memory pool 𝑀. For a query 𝑞, after a plan 𝑝 is selected, the DBMS executes the query 𝑞 according to the plan 𝑝, and the execution time of the plan 𝑝 is 𝑡 (𝑝). The triple 𝐷𝑖 = (𝑞, 𝑝, 𝑡 (𝑝)) constitutes a training instance, and all 𝐷𝑖 are collected in 𝑀. For every 10 executed queries, HybridQO will perform”]; traversing partial data based on the query plan (search strategy) by the hybrid query optimizer using a machine learning executor)
returning a query result based on the uncertainty value and the search strategy ([page 3926, 2.2, HybridQ0 Framework, “Especially, the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty”], [page 3926, col. 1, last paragraph; “we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], [page 3927, col. 1, last paragraph, “we use the execution result to incrementally update the above two models”], returning answer/execution result (a query result) obtained by selecting the optimal query plan (search strategy) and the query plan is selected based the uncertainty values of the query plan).
Yu does not explicitly teach by using a query parser; base data, wherein the supplementary generated data have been generated using Machine Learning.
However, Ren teaches by using a query parser ([page 17, 3.2. Natural Query Engine: “(1) Query parsing to verify syntax correctness]; query engine parses the query (query parser));
traverse at least a part of a data storage and utilize at least a part of an embedding storage ([page; 31, “(2) execute relation traversal in the latent space”]; [page 43, A General Note on Incompleteness: “all the operators should incorporate predicted hard answers in addition to easy answers reachable by graph traversal as in symbolic graph databases”]; [page 52, Entitlement Metrics: “Natural for database querying tasks, it is expected that query answering models first recover easy answers already existing in the graph (reachable by edge traversal)”]; traversing the databases (data storage and embedding storage));
wherein the data storage comprises base data stored in a graph format ([page 17, 1st paragraph “we propose the concept of a neural graph storage (Fig. 7) where both the input graph and its vector representations are sources of truth. Physically, it consists of the (1) Graph Store that stores a (multi-relational) adjacency matrix that in the basic form can be implemented with sparse COO, CSR, or more efficient compressed formats, (2) Feature Store that stores node- and edge-level
feature vectors of various formats and modalities, e.g., numerical, categorical, text data, or already given vectors”], [page 17, last paragraph, “In NGDBs, the sources of truth are both graph adjacency (possibly incomplete) and latent graph representations (e.g., vector representations of entities and relations)”], [page 43, A General Note on Incompleteness: “all the operators should incorporate predicted hard answers in addition to easy answers reachable by graph traversal as in symbolic graph databases”]; graph store stores a multi relational adjacency matrix. An adjacency matrix represents relationships between nodes and entities that are represented in a graph format. Therefore, graph store which is the data storage stores the multi-relational adjacency matrix which is the base data stored in a graph format) and the embedding storage comprises supplementary generated data for at least a part of the base data ([page 17, first paragraph “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, last paragraph, “In NGDBs, the sources of truth are both graph adjacency (possibly incomplete) and latent graph representations (e.g., vector representations of entities and relations)”], [page 17, 1st paragraph “The embedding process can be viewed as a compression step but the semantic and structure similarity of entities/relations is kept. The distance between entities/relations in the embedding space should be positively correlated with the semantic/structure similarity”], [page 4, 4th paragraph “neural graph storage assumes an encoder that further compresses the raw information in a semantic-preserving latent embedding space, i.e., similar entities/relations are mapped to similar embeddings”], the graph store stores the multi-relational adjacency matrix, which corresponds to base data. The embedding store uses an encoder to produce latent graph representations based on the underlying adjacency and associated features. The embedding process is a compression step. Thus, the resulting compressed latent graph representations are the supplementary generated data and are generated based on the underlying adjacency represented by the adjacency matrix (base data)), wherein the supplementary generated data have been generated using Machine Learning ([page 17, first paragraph “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, 2nd paragraph “Another option for the encoder is a graph neural network (GNN) that can be analyzed through the lens of message passing and neighborhood aggregation (Section 2.7)”], [page 17, 3rd paragraph, “The GNN parameters are shared for all the entities and relations on the graph”], the compressed graph representation in the latent space (supplementary data) of the embedding store is generated using encoder/graph neural network (machine learning)).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Ren’s query parser, graph store, embedding store and traversing the data stores into invention of Yu to properly parse incoming queries and traverse data to execute uncertainty-selected query plan over both graphs formatted data and corresponding machine-generated embeddings. Yu and Ren are analogous arts because each art teaches processing queries. The motivation would be to improve the completeness and relevance of query results by retrieving answers through graph traversal while also predicting additional answers from the learned embedding (Ren, [page 24, 3rd paragraph “incorporates types and their hierarchy to improve predictions over entities as well as introduces a task of predicting types of answer entities”]).
With respect to claim 16, Yu and Ren in combination teaches the method according to claim 15, Yu does not explicitly teach wherein the base data are stored in a generic data model supporting graph.
However, Ren teaches wherein the base data are stored in a generic data model supporting graph ([page 16, second paragraph, “The storage layer consists of (1) Graph Store for keeping the multi-relational adjacency matrix in space- and time-efficient manner (e.g., in various sparse formats like COO and CSR”], [page 16, 3.1 Neural Graph Storage, “On a physical level, graph databases store edges employing
various indexes and data structures optimized for read (e.g., B+ trees (Neumann & Weikum, 2008) or HDT (Fernández et al., 2013)) or write applications (such as LSM trees (Sagi et al., 2022))”], [page 17, first paragraph “it consists of the (1) Graph
Store that stores a (multi-relational) adjacency matrix that in the basic form can be implemented with sparse COO, CSR, or more efficient compressed formats”], the graph storage stores basic graph models (generic data model supporting graph)).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Ren’s generic data model supporting graph into invention of Yu to properly store the initial sets of data. Yu and Ren are analogous arts because each art teaches processing queries. The motivation would be to provide a flexible and standardized representation of entities and relationships and facilitate efficient graph traversal and query processing (Ren, [page 24, 3rd paragraph “incorporates types and their hierarchy to improve predictions over entities as well as introduces a task of predicting types of answer entities”]).
With respect to claim 19, Yu and Ren in combination teaches the method according to claim 15, Ren further teaches wherein the supplementary data in the embedding storage are compressed vector representations of all the nodes of the data storage ([page 4, 4th paragraph “neural graph storage assumes an encoder
that further compresses the raw information in a semantic-preserving latent embedding space, i.e., similar entities/relations are mapped to similar embeddings”], [page 17, 1st paragraph “The embedding process can be viewed as a compression step but the semantic and structure similarity of entities/relations is kept. The distance between entities/relations in the embedding space should be positively correlated with
the semantic/structure similarity”], [col. 50, 2nd paragraph “inference is performed on the graph with the same set of nodes and relation types as in training but with different edges”]; the data generated (supplementary data) from the adjacency graph are compressed and stored in the embedding storage for all the nodes).
With respect to claim 21, Yu and Ren in combination teaches the method according to claim 15, Yu does not explicitly teach wherein the query has any graph query language format.
However, Ren teaches wherein the query has any graph query language format ([page 4, last paragraph “This includes definitions of different types of graphs,
graph queries and their mapping to structured languages like SPARQL, formalizing approximate graph query answering, logical operators, and fuzzy logic”], [page 27, 1st paragraph, “The shape of the entity embedding matrix is |E| × d (|R| × d), where d is the dimension of the embedding. Shallow embedding methods assume independence of the representation of all the nodes on the graph”], [col. 7, last paragraphs “These closely resemble our basic graph query for RDF graphs”], the query has SPARQL/RDF language which is a graph query language format).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Ren’s graph query format language into invention of Yu to process query faster. Yu and Ren are analogous arts because each art teaches processing queries. The motivation would be to efficiently formulate and execute queries to save time and cost of a system (Ren, [page 24, 3rd paragraph “incorporates types and their hierarchy to improve predictions over entities as well as introduces a task of predicting types of answer entities”]).
With respect to claim 23, Yu and Ren in combination teaches the method according to claim 15, Yu further teaches wherein a cost model is included in the search strategy of the hybrid query optimizer, which reflects the time needed for the method to provide a query under a given uncertainty value ([page 3926, col. 1, 2.2 HybridQ0 Framework, “Our basic idea is to combine the learning-based optimizer and the cost-based optimizer to select a good plan”], [page 3929, col. 1, 4. Optimal plan selection, “we leverage the uncertainty-aware plan performance predictor to generate the estimated execution time 𝑡𝑖 and the corresponding uncertainty 𝑢𝑖 for each candidate plan 𝑝𝑖. Then we select the optimal plan from these candidates according to 𝑡𝑖 and 𝑢𝑖”], [page 3930, col. 1, 4.4. Uncertainty-based Optimal Plan Selection, “MHPE outputs the execution time and uncertainty (𝑇(𝑝),𝑈 (𝑝)) = 𝑓𝜃 (𝐸𝑄,𝑅𝑝) for each plan 𝑝. We need to choose the best plan to give to the DBMS for the final execution”], time and cost for the search strategy for each uncertainty value).
With respect to claim 24, Yu and Ren in combination teaches the method according to claim 15, Yu further teaches wherein the hybrid query optimizer enumerates candidate hybrid plans, which find the optimal cost for a given uncertainty value for a specific query ([page 3926, col. 1, 2.2 HybridQ0 Framework, “Our basic idea is to combine the learning-based optimizer and the cost-based optimizer to select a good plan”], [page 3929, col. 1, 4. Optimal plan selection, “we leverage the uncertainty-aware plan performance predictor to generate the estimated execution time 𝑡𝑖 and the corresponding uncertainty 𝑢𝑖 for each candidate plan 𝑝𝑖. Then we select the optimal plan from these candidates according to 𝑡𝑖 and 𝑢𝑖”], [page 3930, col. 1, 4.4. Uncertainty-based Optimal Plan Selection, “MHPE outputs the execution time and uncertainty (𝑇(𝑝),𝑈 (𝑝)) = 𝑓𝜃 (𝐸𝑄,𝑅𝑝) for each plan 𝑝. We need to choose the best plan to give to the DBMS for the final execution”], time and cost for a specific query for each uncertainty value).
With respect to claim 26, Yu and Ren in combination teaches the method according to claim 15, Yu does not explicitly teach wherein the query comprises a plurality of operators that execute on a graph.
However, Ren teaches wherein the query comprises a plurality of operators that execute on a graph ([page 19, 6th paragraph, “executing logical operators as intermediate steps; (2) global, when the entire query graph is encoded and executed in a latent space in one step. For example (Fig. 7), given a query plan q = U?.9V : win(TuringAward, V ) ^ field(DeepLearning, V ) ^ university(V,U?),
the atomic mechanism executes separate relation projections, e.g., win(TuringAward, V ), sequentially while the global mechanism encodes the whole query graph and probes it against the latent space of the graph”], logical operators).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Ren’s plurality of operators that execute on a graph into invention of Yu to process query faster. Yu and Ren are analogous arts because each art teaches processing queries. The motivation would be to efficiently formulate and execute queries to save time and cost of a system (Ren, [page 24, 3rd paragraph “incorporates types and their hierarchy to improve predictions over entities as well as introduces a task of predicting types of answer entities”]).
With respect to claim 29, Yu and Ren in combination teaches the method according to claim 15, Yu does not explicitly teach wherein the generic data model supporting graph is a property graph data model or a relational model.
However, Ren teaches wherein the generic data model supporting graph is a property graph data model or a relational model ([page 21, 2nd paragraph “hyper-relational KGs following the RDF-star or Labeled Property Graph (LPG) Formats”], LPG is a property graph data model).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Ren’s property graph data model into invention of Yu to process query faster. Yu and Ren are analogous arts because each art teaches processing queries. The motivation would be to efficiently formulate and execute queries to save time and cost of a system (Ren, [page 24, 3rd paragraph “incorporates types and their hierarchy to improve predictions over entities as well as introduces a task of predicting types of answer entities”]).
With respect to claim 30, Yu and Ren in combination teaches the method according to claim 15, Yu further teaches wherein the uncertainty value is provided by a user or by the hybrid query optimizer ([page 3929, col. 2, 4.3 Multi-head Performance Estimator, “uncertainty [15]. The uncertainty output by the neural network is composed of Epistemic Uncertainty and Aleatoric Uncertainty [15, 18]. (1) Epistemic Uncertainty𝑈𝐸 is caused by the fact that the parameters of the neural network model are not good enough], the uncertainty value is provided by the hybrid query optimizer).
With respect to claim 31, Yu and Ren combination teach the method according to claim 15, Yu further teaches the comprising the step of choosing one or more search strategies for processing the query by utilizing a hybrid query optimizer, wherein each search strategy is associated with an uncertainty value ([page 3926, 2.2, HybridQ0 Framework, “Especially, the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty”], each search query plan is associated with a uncertainty value and the query plan is processed by using a hybrid query optimizer), and returning a query result and an uncertainty value for each search strategy ([page 3926, col. 1, last paragraph; “we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], [page 3927, col. 1, last paragraph, “we use the execution result to incrementally update the above two models”], ([page 3926, 2.2, HybridQ0 Framework, “Especially, the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty”], [page 3926, col. 1, last paragraph; “we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], [page 3927, col. 1, last paragraph, “we use the execution result to incrementally update the above two models”], [page 3930, 1st paragraph, “MHPE outputs the execution time and uncertainty (𝑇 (𝑝),𝑈 (𝑝)) =𝑓𝜃 (𝐸𝑄, 𝑅𝑝 ) for each plan 𝑝. We need to choose the best plan to give to the DBMS for the final execution”], returning uncertainty value p for each query plan (search strategy) and also returning the answers/execution results (query results)).
With respect to claim 32, Yu teaches a computer implemented method for handling queries, the method comprising the steps of ([page 3926, col. 1, element: 2.2 Hybrid00 Framework; “we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models. As shown in Figure 2, the entire framework consists of three key modules”]; selecting queries to process (handling queries)):
reading the query ([page 3926, col. 1, element 2.2, “we leverage a learning-based uncertainty model to select the best plan from candidate plans. Finally, we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], receiving the query and selecting the best plan to execute the query (reading the query));
choosing one or more search strategies for processing the query by utilizing a hybrid query optimizer, wherein each search strategy is associated with an uncertainty value ([page 3925, col. 1, last paragraph “We propose an adaptive hybrid optimizer that combines the cost based and learning-based optimizer through the hint functionality”], [page 3926, col. 1, element 2.2, “Hybrid00 Framework”; “the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty]; [page 3930, col. 1, element 4.4, Uncertainty-based Optimal Plan Selection; “We first filter out the learning-based plans with high uncertainty to obtain the remaining 𝐻 accurately estimated plans, each of which is denoted”], [page 3930, col. 2, “then use the mapping to keep (drop) low (high) uncertainty queries”], selecting the best plan (search strategy) with low uncertainty (uncertainty value) and the candidate plans are generated and evaluated using a hybrid query optimizer);
based on each search strategy of the hybrid query optimizer, using a Machine Learning-based query executor to traverse at least a part of a data ((page 3926, col. 1, element 2.2, “As it is expensive to fully traverse the leading hint tree, we propose a learning-based leading hint selection method 𝑖.𝑒., MCTS, to select leading hints. Then, we use the hint functionality in DBMS to generate complete plans as candidate plans”], [page 3926, col. 1, element 2.2, “query. Then we first select leading hints from the leading hint tree by MCTS which is a typical tree search algorithm. We will discuss how to use MCTS to select leading hints later. For each selected leading hint, 𝑖.𝑒., the left-deep join orders of the prefixes, we leverage the cost-based optimizer in DBMS to produce a complete query plan. For each hint, we generate a candidate plan”], [page 3930, col. 2, element 4.5, “As shown in Figure 2, HybridQO collects all training data with a memory pool 𝑀. For a query 𝑞, after a plan 𝑝 is selected, the DBMS executes the query 𝑞 according to the plan 𝑝, and the execution time of the plan 𝑝 is 𝑡 (𝑝). The triple 𝐷𝑖 = (𝑞, 𝑝, 𝑡 (𝑝)) constitutes a training instance, and all 𝐷𝑖 are collected in 𝑀. For every 10 executed queries, HybridQO will perform”]; traversing partial data based on the query plan (search strategy) by the hybrid query optimizer using a machine learning executor)
and returning a query result and an uncertainty value for each search strategy ([page 3926, 2.2, HybridQ0 Framework, “Especially, the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty”], [page 3926, col. 1, last paragraph; “we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], [page 3927, col. 1, last paragraph, “we use the execution result to incrementally update the above two models”], [page 3930, 1st paragraph, “MHPE outputs the execution time and uncertainty (𝑇 (𝑝),𝑈 (𝑝)) =𝑓𝜃 (𝐸𝑄, 𝑅𝑝 ) for each plan 𝑝. We need to choose the best plan to give to the DBMS for the final execution”], returning uncertainty value p for each query plan (search strategy) and also returning the answers/execution results (query results)).
Yu does not explicitly teach by using a query parser;
However, Ren teaches by using a query parser ([page 17, 3.2. Natural Query Engine: “(1) Query parsing to verify syntax correctness]; query engine parses the query (query parser));
to traverse at least a part of a data storage and utilize at least a part of an embedding storage ([page; 31, “(2) execute relation traversal in the latent space”]; [page 43, A General Note on Incompleteness: “all the operators should incorporate predicted hard answers in addition to easy answers reachable by graph traversal as in symbolic graph databases”]; [page 52, Entitlement Metrics: “Natural for database querying tasks, it is expected that query answering models first recover easy answers already existing in the graph (reachable by edge traversal)”]; traversing graph databases (data storage)); wherein the data storage comprises base data stored in a graph format ([page 17, 1st paragraph “we propose the concept of a neural graph storage (Fig. 7) where both the input graph and its vector representations are sources of truth. Physically, it consists of the (1) Graph Store that stores a (multi-relational) adjacency matrix that in the basic form can be implemented with sparse COO, CSR, or more efficient compressed formats, (2) Feature Store that stores node- and edge-level feature vectors of various formats and modalities, e.g., numerical, categorical, text data, or already given vectors”], [page 17, last paragraph, “In NGDBs, the sources of truth are both graph adjacency (possibly incomplete) and latent graph representations (e.g., vector representations of entities and relations)”], [page 43, A General Note on Incompleteness: “all the operators should incorporate predicted hard answers in addition to easy answers reachable by graph traversal as in symbolic graph databases”]; graph store stores a multi relational adjacency matrix. An adjacency matrix represents relationships between nodes and entities that are represented in a graph format. Therefore, graph store which is the data storage stores the multi-relational adjacency matrix which is the base data stored in a graph format) and the embedding storage comprises supplementary generated data for at least a part of the base data ([page 17, first paragraph “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, last paragraph, “In NGDBs, the sources of truth are both graph adjacency (possibly incomplete) and latent graph representations (e.g., vector representations of entities and relations)”], [page 17, 1st paragraph “The embedding process can be viewed as a compression step but the semantic and structure similarity of entities/relations is kept. The distance between entities/relations in the embedding space should be positively correlated with the semantic/structure similarity”], [page 4, 4th paragraph “neural graph storage assumes an encoder that further compresses the raw information in a semantic-preserving latent embedding space, i.e., similar entities/relations are mapped to similar embeddings”], the graph store stores the multi-relational adjacency matrix, which corresponds to base data. The embedding store uses an encoder to produce latent graph representations based on the underlying adjacency and associated features. The embedding process is a compression step. Thus, the resulting compressed latent graph representations are the supplementary generated data and are generated based on the underlying adjacency represented by the adjacency matrix (base data)), wherein the supplementary generated data have been generated using Machine Learning ([page 17, first paragraph “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, 2nd paragraph “Another option for the encoder is a graph neural network (GNN) that can be analyzed through the lens of message passing and neighborhood aggregation (Section 2.7)”], [page 17, 3rd paragraph, “The GNN parameters are shared for all the entities and relations on the graph”], the compressed graph representation in the latent space (supplementary data) of the embedding store is generated using encoder/graph neural network (machine learning)).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Ren’s query parser, graph store, embedding store and traversing the data stores into invention of Yu to properly parse incoming queries and traverse data to execute uncertainty-selected query plan over both graphs formatted data and corresponding machine-generated embeddings. Yu and Ren are analogous arts because each art teaches processing queries. The motivation would be to improve the completeness and relevance of query results by retrieving answers through graph traversal while also predicting additional answers from the learned embedding (Ren, [page 24, 3rd paragraph “incorporates types and their hierarchy to improve predictions over entities as well as introduces a task of predicting types of answer entities”]).
With respect to claim 33, Yu teaches a query handling system comprising:
a server unit for storing data; and a processing unit configured to ((Yu, page 3929, 1st paragraph, “Then we apply LSTM to encode the join order, which consists of several steps. At 𝑖-th step, the hidden state vector ℎ𝑖 represents current state and the memory cell𝑚𝑖 preserves the information over (𝑒 (𝑡1), 𝑒 (𝑡2), . . . , 𝑒 (𝑡𝑖 )). LSTM uses an LSTMUnit to get the state ℎ𝑖 and the memory 𝑚𝑖 based on the input 𝑒 (𝑡𝑖 ) and the previous state representation”], Ren page 18, last paragraph “very large graphs that cannot be stored entirely in the main memory”], [page 3931, 1st col. 1, 5.1 Experimental Setup “The server is a machine with 64 GB of RAM, a 4.00 GHz i9 CPU,
and an NVIDIA 1080ti GPU. The neural networks are implemented using Pytorch”], computer contains memory and processor and server)):
obtain a query ([page 3926, col. 1, element 2.2, “we leverage a learning-based uncertainty model to select the best plan from candidate plans. Finally, we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], receiving the query (obtain a query) and selecting the best plan to execute the query (reading the query));
choose an optimal way of processing the query by utilizing an uncertainty value provided and a hybrid query optimizer to provide a search strategy ([page 3925, col. 1, last paragraph “We propose an adaptive hybrid optimizer that combines the cost based and learning-based optimizer through the hint functionality”], [page 3926, col. 1, element 2.2, “Hybrid00 Framework”; “the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty]; [page 3930, col. 1, element 4.4, Uncertainty-based Optimal Plan Selection; “We first filter out the learning-based plans with high uncertainty to obtain the remaining 𝐻 accurately estimated plans, each of which is denoted”], [page 3930, col. 2, “then use the mapping to keep (drop) low (high) uncertainty queries”], selecting the best plan (search strategy) with low uncertainty (uncertainty value) and the candidate plans are generated and evaluated using a hybrid query optimizer);
based on the search strategy of the hybrid query optimizer, use a Machine Learning- based query executor to traverse a part of a data ((page 3926, col. 1, element 2.2, “As it is expensive to fully traverse the leading hint tree, we propose a learning-based leading hint selection method 𝑖.𝑒., MCTS, to select leading hints. Then, we use the hint functionality in DBMS to generate complete plans as candidate plans”], [page 3926, col. 1, element 2.2, “query. Then we first select leading hints from the leading hint tree by MCTS which is a typical tree search algorithm. We will discuss how to use MCTS to select leading hints later. For each selected leading hint, 𝑖.𝑒., the left-deep join orders of the prefixes, we leverage the cost-based optimizer in DBMS to produce a complete query plan. For each hint, we generate a candidate plan”], [page 3930, col. 2, element 4.5, “As shown in Figure 2, HybridQO collects all training data with a memory pool 𝑀. For a query 𝑞, after a plan 𝑝 is selected, the DBMS executes the query 𝑞 according to the plan 𝑝, and the execution time of the plan 𝑝 is 𝑡 (𝑝). The triple 𝐷𝑖 = (𝑞, 𝑝, 𝑡 (𝑝)) constitutes a training instance, and all 𝐷𝑖 are collected in 𝑀. For every 10 executed queries, HybridQO will perform”]; traversing partial data based on the query plan (search strategy) by the hybrid query optimizer using a machine learning executor)
and return a query result based on the uncertainty value and the search strategy ([page 3926, 2.2, HybridQ0 Framework, “Especially, the model not only predicts the query performance, but also computes the uncertainty of the prediction. The uncertainty measures how confident is the model for the prediction. Therefore, intuitively, we should select the optimal plan with high performance as well as low uncertainty”], [page 3926, col. 1, last paragraph; “we run the selected plan to answer the query and use this plan (and its running time) as a training example to update the two learning models”], [page 3927, col. 1, last paragraph, “we use the execution result to incrementally update the above two models”], returning answer/execution result (a query result) obtained by selecting the best query plan (search strategy) and the query plan is selected based the uncertainty values of the query plan).
Yu does not explicitly teach
However, Ren teaches traverse at least a part of a data storage and utilize at least a part of an embedding storage ([page; 31, “(2) execute relation traversal in the latent space”]; [page 43, A General Note on Incompleteness: “all the operators should incorporate predicted hard answers in addition to easy answers reachable by graph traversal as in symbolic graph databases”]; [page 52, Entitlement Metrics: “Natural for database querying tasks, it is expected that query answering models first recover easy answers already existing in the graph (reachable by edge traversal)”]; traversing graph databases (data storage));
wherein the data storage comprises base data stored in a graph format ([page 17, 1st paragraph “we propose the concept of a neural graph storage (Fig. 7) where both the input graph and its vector representations are sources of truth. Physically, it consists of the (1) Graph Store that stores a (multi-relational) adjacency matrix that in the basic form can be implemented with sparse COO, CSR, or more efficient compressed formats, (2) Feature Store that stores node- and edge-level
feature vectors of various formats and modalities, e.g., numerical, categorical, text data, or already given vectors”], [page 17, last paragraph, “In NGDBs, the sources of truth are both graph adjacency (possibly incomplete) and latent graph representations (e.g., vector representations of entities and relations)”], [page 43, A General Note on Incompleteness: “all the operators should incorporate predicted hard answers in addition to easy answers reachable by graph traversal as in symbolic graph databases”]; graph store stores a multi relational adjacency matrix. An adjacency matrix represents relationships between nodes and entities that are represented in a graph format. Therefore, graph store which is the data storage stores the multi-relational adjacency matrix which is the base data stored in a graph format) and the embedding storage comprises supplementary generated data for at least a part of the base data ([page 17, first paragraph “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, last paragraph, “In NGDBs, the sources of truth are both graph adjacency (possibly incomplete) and latent graph representations (e.g., vector representations of entities and relations)”], [page 17, 1st paragraph “The embedding process can be viewed as a compression step but the semantic and structure similarity of entities/relations is kept. The distance between entities/relations in the embedding space should be positively correlated with the semantic/structure similarity”], [page 4, 4th paragraph “neural graph storage assumes an encoder that further compresses the raw information in a semantic-preserving latent embedding space, i.e., similar entities/relations are mapped to similar embeddings”], the graph store stores the multi-relational adjacency matrix, which corresponds to base data. The embedding store uses an encoder to produce latent graph representations based on the underlying adjacency and associated features. The embedding process is a compression step. Thus, the resulting compressed latent graph representations are the supplementary generated data and are generated based on the underlying adjacency represented by the adjacency matrix (base data)), wherein the supplementary generated data have been generated using Machine Learning ([page 17, first paragraph “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, 2nd paragraph “Another option for the encoder is a graph neural network (GNN) that can be analyzed through the lens of message passing and neighborhood aggregation (Section 2.7)”], [page 17, 3rd paragraph, “The GNN parameters are shared for all the entities and relations on the graph”], the compressed graph representation in the latent space (supplementary data) of the embedding store is generated using encoder/graph neural network (machine learning)).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Ren’s query parser, graph store, embedding store and traversing the data stores into invention of Yu to properly parse incoming queries and traverse data to execute uncertainty-selected query plan over both graphs formatted data and corresponding machine-generated embeddings. Yu and Ren are analogous arts because each art teaches processing queries. The motivation would be to improve the completeness and relevance of query results by retrieving answers through graph traversal while also predicting additional answers from the learned embedding (Ren, [page 24, 3rd paragraph “incorporates types and their hierarchy to improve predictions over entities as well as introduces a task of predicting types of answer entities”]).
Claim 34 encompasses the same scope of limitations of claim 15, in addition of a non-transitory computer-readable medium storing instruction that, when executed by one or more processors, cause the one or more processors to carry out the method of handling queries (Yu, page 3929, 1st paragraph, “Then we apply LSTM to encode the join order, which consists of several steps. At 𝑖-th step, the hidden state vector ℎ𝑖 represents current state and the memory cell𝑚𝑖 preserves the information over (𝑒 (𝑡1), 𝑒 (𝑡2), . . . , 𝑒 (𝑡𝑖 )). LSTM uses an LSTMUnit to get the state ℎ𝑖 and the memory 𝑚𝑖 based on the input 𝑒 (𝑡𝑖 ) and the previous state representation”], Ren page 18, last paragraph “very large graphs that cannot be stored entirely in the main memory”], computer contains memory and processor and non-transitory computer readable medium). Therefore, claim 34 is rejected based on the same basis of rejection of claim 15.
Claim(s) 17, 25 is/are rejected under 35 U.S.C. 103 as being unpatentable over, NPL “Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection”, by Xiang Yu filed on September 1, 2022, pages 3924-3936 and in view of “Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases” by Hongyu Ren filed on March 26, 2023, pages 1-65 and in view of Kulkarni et al. (US 2023/0081891).
With respect to claim 17, Yu and Ren in combination teach the method according to claim 15, but do not in combination explicitly teach wherein the query result is stored as new data in incomplete graphs of the data storage.
However, Kulkarni teaches wherein the query result is stored as new data in incomplete graphs of the data storage ([0004, information extracted from sources like documents and databases are populated into knowledge graphs], [0082, This requires populating and updating the missing links from the relevant answers received from experts in response to the additionally informative questions], [0093, the knowledge graphs are identified for missing links such that inquisitive and contextually relevant questions may be asked], the answers (query result) are populated (stored as new data) in the missing links of the graph (incomplete graph), the graph is stored in the data storage).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Kulkarni’s storing query results as new data in the incomplete graphs of the data storage into invention of Yu/Ren to store new data and complete the graphs. Yu, Ren, Kulkarni are analogous arts because each art teaches processing queries. The motivation would be to improve the completeness and relevance of query results by retrieving answers through graph traversal while also predicting additional answers from the learned embedding (Ren, [0043, illustrative embodiments improve the performance and operation of knowledge management within knowledge graphs by identifying gaps persisting in the available knowledge graph and correcting/updating/completing the graph with additional knowledge customarily held within the minds of subject matter experts]).
With respect to claim 25, Yu and Ren in combination teach the method according to claim 15, Ren teaches wherein the embedding storage comprises ([page 17, first paragraph “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, 2nd paragraph “Another option for the encoder is a graph neural network (GNN) that can be analyzed through the lens of message passing and neighborhood aggregation (Section 2.7)”], embedding storage storing supplementary data) but does not explicitly teach the embedding storage comprises supplementary generated data based on incomplete graphs of the data storage.
However, Kulkarni teaches supplementary generated data based on incomplete graphs of the data storage ([0004, information extracted from sources like documents and databases are populated into knowledge graphs], [0082, This requires populating and updating the missing links from the relevant answers received from experts in response to the additionally informative questions], [0093, the knowledge graphs are identified for missing links such that inquisitive and contextually relevant questions may be asked], the data are generated based on the missing links (incomplete) of the graph (incomplete graph), the graph is stored in the data storage).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Kulkarni’s supplementary generated data in the incomplete graphs of the data storage into invention of Yu/Ren to store data and complete the graphs. Yu, Ren, Kulkarni are analogous arts because each art teaches processing queries. The motivation would be to improve the completeness and relevance of query results by retrieving answers through graph traversal while also predicting additional answers from the learned embedding (Kulkarni, [0043, illustrative embodiments improve the performance and operation of knowledge management within knowledge graphs by identifying gaps persisting in the available knowledge graph and correcting/updating/completing the graph with additional knowledge customarily held within the minds of subject matter experts]).
Claim(s) 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over, NPL “Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection”, by Xiang Yu filed on September 1, 2022, pages 3924-3936 and in view of “Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases” by Hongyu Ren filed on March 26, 2023, pages 1-65 and in view of Alpers et al. (US 2014/0214797).
With respect to claim 18, Yu and Ren in combination teach the method according to claim 15, but do not in combination explicitly teach wherein the uncertainty value comprises a global uncertainty value set at a global level of the query, and/or a local uncertainty value, set at a local level of the query.
However, Alpers teaches wherein the uncertainty value comprises a global uncertainty value set at a global level of the query, and/or a local uncertainty value, set at a local level of the query ([0068, An example would be a query with the following predicate: "AGE<?". This type of predicate typically has high uncertainty because the value of the parameter marker "?" is not known until execution time. The predicate may qualify all rows or no rows, and the query optimizer does not know which will occur at execution time (hence, there is high uncertainty], uncertainty value is higher for the specific predicate in the query (local level of the query), the uncertainty value is set to the local level of the query).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Alpers’s local level uncertainty value into invention of Yu/Ren to have uncertainty values for specific part of the queries. Yu, Ren, Alpers are analogous arts because each art teaches processing queries. The motivation would be to improve the accuracy and efficiency of the queries to select the best query for faster query processing (Alpers, [0014, improve the accuracy of the cost-estimates made by the query optimizer over many executions of the query.]).
Claim(s) 20 is/are rejected under 35 U.S.C. 103 as being unpatentable over, NP “Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection”, by Xiang Yu filed on September 1, 2022, pages 3924-3936 and in view of “Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases” by Hongyu Ren filed on March 26, 2023, pages 1-65 and in view of Pai et al. (US 2024/0127909).
With respect to claim 20, Yu and Ren in combination teach the method according to claim 15, but do not in combination explicitly teach wherein the embedding storage is continuously supplemented with additional supplementary data based on the data storage.
However, Pai teaches wherein the embedding storage is continuously supplemented with additional supplementary data based on the data storage ([0016, the training system may continuously receive the initial knowledge graph from the data structure, may periodically receive the initial knowledge graph from the data structure], [0023, the training system may continuously receive the new knowledge graph from the data structure], the training system continuously receives knowledge graph from the data structure; Ren teaches receiving data in the embedding storage based on the data storage and Pai teaches continuously receiving data).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Pai’s continuously receiving data into invention of Yu/Ren to store new data. Yu, Ren, Pai are analogous arts because each art teaches processing data. The motivation would be to improve efficiency while preserving accurate retrieval of newly added graph information (Pai, [0012, fully retraining the KGE model when new information not associated with existing information of the knowledge graph is added to the knowledge graph, handling large datasets represented by knowledge graphs, updating large knowledge graphs, and/or the like.]).
Claim(s) 22 is/are rejected under 35 U.S.C. 103 as being unpatentable over, NP “Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection”, by Xiang Yu filed on September 1, 2022, pages 3924-3936 and in view of “Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases” by Hongyu Ren filed on March 26, 2023, pages 1-65 and in view of Clark et al. (US 2016/0110357).
With respect to claim 22, Yu and Ren in combination teach the method according to claim 15, Ren teaches the computer implemented method on the data storage and on the embedding storage ([page 17, 1st paragraph; “Physically, it consists of the (1) Graph Store that stores a (multi-relational) adjacency matrix that in the basic form can be implemented with sparse COO, CSR, or more efficient compressed formats, “(3) Embedding Store that leverages an Encoder to produce graph representations in a latent space based on the underlying adjacency and associated features”], [page 17, 2nd paragraph “Another option for the encoder is a graph neural network (GNN) that can be analyzed through the lens of message passing and neighborhood aggregation (Section 2.7)”], the embedding data storage and data storage).
Yu and Ren do not explicitly teach wherein the uncertainty value determines the amount of processes performed by the computer implemented method on the data storage and on the embedding storage.
However, Clark teaches wherein the uncertainty value determines the amount of processes performed by the computer implemented method on the data storage and on the embedding storage ([0054, each confidence score may indicate a level of confidence in a given candidate answer based on a number of factors. These factors may include, for example, the trustworthiness of a source item's publisher, the content of a source item], [0055, Different threshold confidence scores may be used under different circumstances. If at least one of the confidence scores is above the threshold level of confidence, then, per block 310, one or more of the candidate answers (e.g., the candidate answer with the highest confidence score) may be presented to the user], when the confidence scores (uncertainty score) goes below a threshold or goes above a threshold it determines amount of processes because when there is sufficient amount the processing terminates and Ren teaches data storage and embedding storage).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Clark’s uncertainty value determining amount of processed into invention of Yu/Ren to determine amount of processes. Yu, Ren, Clark are analogous arts because each art teaches processing data. The motivation would be to improve the accuracy and reality of the query to have a robust system (Clark, [0018, there may be some advantage in being able to predict the situations in which these searches should be performed]).
Claim(s) 27 is/are rejected under 35 U.S.C. 103 as being unpatentable over, NP “Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection”, by Xiang Yu filed on September 1, 2022, pages 3924-3936 and in view of “Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases” by Hongyu Ren filed on March 26, 2023 pages 1-65 and in view of Najork et al. (US -2011/0202846).
With respect to claim 27, Yu and Ren in combination teach the method according to claim 15, but do not in combination teach wherein the plurality of operators that execute on a graph are node distance approximates.
However, Najork teaches wherein the plurality of operators that execute on a graph are node distance approximates ([0017, The estimated shortest distance may then be returned to the user at the computing device 100], [0027, The shortest distance service 106 of the shortest distance server 108 may use the generated sketches to estimate the shortest distance between a first node and a second node of a graph represented by the graph data 115], the estimated shortest distance between nodes are identified, operators are query operations i.e. retrieving, summing, finding distances etc., moreover, Ren teaches operator page 4, 2nd paragraph “operators with learnable vector representations and neural networks”).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Najork’ s node distance approximates into invention of Yu/Ren to process the query faster. Yu, Ren, Najork are analogous arts because each art teaches processing data. The motivation would be to reduce query processing time while still producing a sufficiently accurate distance result, particularly for large graphs (Najork, [0025, the value of K selected may be determined based on the number of nodes in the graph data 115, and the available computing resources or desired accuracy, for example]).
Claim(s) 28 is/are rejected under 35 U.S.C. 103 as being unpatentable over, NP “Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection”, by Xiang Yu filed on September 1, 2022, pages 3924-3936 and in view of “Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases” by Hongyu Ren filed on March 26, 2023, pages 1-65 and in view of Beskales et al. (US 11,416,780).
With respect to claim 28, Yu and Ren in combination teach the method according to claim 15, but do not in combination teach wherein the uncertainty value has limits from 0 to 1, and it is less than 0.5, or less than 0.35, or less than 0.25, or around 0.05 and 0.15, or around 0.1.
However, Beskales teaches wherein the uncertainty value has limits from 0 to 1, and it is less than 0.5, or less than 0.35, or less than 0.25, or around 0.05 and 0.15, or around 0.1 ([col. 6, lines 55-60, “pairwise similarity is in the range [0, 1] and clusters are formed using the Weighted Pair Group Method with Arithmetic Mean (WPGMA) with an average similarity cutoff threshold of 0.5. This ensures that the intra-cluster confidence of each cluster will be at least 0.5, and at most 1”], [col. 6, lines 61-66, “One can convert the confidence metric to an uncertainty metric by subtracting from 1; since intra-cluster confidence is in the range [0.5, 1], 1-intra-cluster confidence will be in the range [0, 0.5], with higher confidence translating to lower uncertainty. For singleton clusters, intra-cluster confidence of 1 will translate to an intra-cluster uncertainty of 0.”], converting the intra-cluster confidence metric into an uncertainty metric by subtracting the confidence value from 1. Because confidence is in the range [0.5, 1], therefore, providing the uncertainty values less than 0.5 and with the limits from 0-1).
Therefore, it would have been obvious to one of the ordinary skills in the art before the effective filing date to incorporate Beskales’s uncertainty value being less than 0.5 and ranges between 0 to 1 into invention of Yu/Ren to provide greater confidence in the clustering results. Yu, Ren, Beskales are analogous arts because each art teaches processing data. The motivation would be to improve the reliability of the system to process query faster (Beskales, [col. 7, lines 7-9, “eliminating the costly operation of identifying all pairs for each cluster”]).
Prior arts Made of Record
The Prior arts made of record and not relied upon are considered pertinent to the applicant’s disclosure.
Morris et al. (US 2019/0197163) teaches identifying query plan and cost of the query plan in paragraphs [0017, 0019, 0022].
Lee et al. (US 2024/0086404) teaches query plans in paragraphs [0020, 0022].
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to FATIMA P MINA whose telephone number is (571)270-3556. The examiner can normally be reached Monday - Friday 9:00 am - 5:00 pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Ann Lo can be reached at 571-272-9767. 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.
/FATIMA P MINA/Examiner, Art Unit 2159 /MARC S SOMERS/Primary Examiner, Art Unit 2159
