# US 12,111,798 B2 · Cloud based schema mapping feature store > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,111,798 B2 - Document ID: US12111798B2 - Status: Granted - Kind: B2 - Application number: 17/587,346 - Filed: 2022-01-28 - Published: 2023-08-03 - Granted: 2024-10-08 - Earliest claimed: 2022-01-28 - Assignees: Oracle International Corporation - Applicants: Oracle International Corporation - Inventors: Yagnesh Dilipbhai Kotecha; Hari Bhaskar Sankaranarayanan; Sandeep Jain; Jagathi Harshitha Arumalla - Web page: https://patentgrep.com/patent/US12111798B2 ## CPC - G06F 16/212: {with details for data modelling support} - G06F 40/295: Named entity recognition - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Embodiments map a source schema to a target schema using a feature store. Embodiments receive a file including a plurality of source schema elements and a plurality of target schema elements, the file including a plurality of unmapped elements. Embodiments retrieve rule based mappings for the unmapped elements between the source schema elements and the target schema elements. Based on semantic matching of the source schema elements, embodiments retrieve feature store based mappings from the feature store for the unmapped elements between the source schema elements and the target schema elements. Embodiments then generate one or more similarity scores for mappings of the source schema elements to the target schema elements. ## Claims ### Claim 1 A method of mapping a source schema of a first cloud based application to a target schema of a second cloud based application using a cloud based feature store, wherein the feature store comprises a central repository of features used for machine learning model training, the method comprising: receiving a file comprising a plurality of source schema terms of the first application and a plurality of target schema terms of the second application, the file comprising a plurality of unmapped terms, wherein the source schema specifies how each of the source schema terms provide a first application name for each underlying data element of first application data stored in a first database, and the target schema specifies how the each of the target schema terms provide a second application name for each underlying data element of second application data stored in a second database, wherein the underlying data elements stored in the first database are the same as the underlying data elements stored in the second database; parsing the source schema and the target schema and extracting source schema terms and target schema terms that correspond to the underlying data elements; retrieving rule based mappings for the unmapped terms between the source schema terms and the target schema terms and generating rule based matches between source schema terms and target schema terms; based on semantic matching of the source schema terms, retrieving feature store based mappings from the feature store for the unmapped terms between the source schema terms and the target schema terms that were not matched using the rule based mappings, wherein the semantic matching comprises classifying named entities mentioned in unstructured text into pre-defined categories using a trained named entity recognition (NER) model; generating one or more similarity scores for the feature store based mappings of the source schema terms to the target schema terms; and generating possible feature store based mappings for the extracted source schema terms and target schema terms based on the similarity scores; wherein the feature store comprises key-value pairs with keys corresponding to source schema terms and values corresponding to matched target schema terms; transferring the underlying data elements from the first database to the second database and, based on the feature store based mappings, automatically transforming by the feature store the source schema terms to corresponding target schema terms for each of the underlying data elements, the transforming comprising converting source messages of the source schema into target messages of the target schema. ### Claim 2 (depends on 1) The method of claim 1, the generating one or more similarity scores comprises using the trained named NER model. ### Claim 3 (depends on 2) The method of claim 2, further comprising: receiving one or more manual mappings of the source schema terms to the target schema terms in response to the possible feature store based mappings; re-training the trained NER model based on the manual mappings. ### Claim 4 (depends on 3) The method of claim 3, further comprising updating the feature store with the re-trained NER model. ### Claim 5 (depends on 1) The method of claim 1, wherein the retrieving feature store based mappings from the feature store for the unmapped terms occurs after the retrieving rule based mappings for any remaining unmapped terms. ### Claim 6 (depends on 5) The method of claim 5, wherein the generating one or more similarity scores for mappings occurs after the retrieving feature store based mappings from the feature store for any remaining unmapped terms. ### Claim 7 (depends on 1) The method of claim 1, wherein the source schema and target schema both comprise an Extensible Markup Language (XML) schema or a JavaScript Object Notation (JSON) schema and the keys in the feature store are organized based on domain, sub domain, entities and values. ### Claim 8 A non-transitory computer readable medium having instructions stored thereon that, when executed by one or more processors on a cloud infrastructure, cause the processors to map a source schema of a first cloud based application to a target schema of a second cloud based application using a cloud based feature store, wherein the feature store comprises a central repository of features used for machine learning model training, the mapping comprising: receiving a file comprising a plurality of source schema terms of the first application and a plurality of target schema terms of the second application, the file comprising a plurality of unmapped terms, wherein the source schema specifies how each of the source schema terms provide a first application name for each underlying data element of first application data stored in a first database, and the target schema specifies how the each of the target schema terms provide a second application name for each underlying data element of second application data stored in a second database, wherein the underlying data elements stored in the first database are the same as the underlying data elements stored in the second database; parsing the source schema and the target schema and extracting source schema terms and target schema terms that correspond to the underlying data elements; retrieving rule based mappings for the unmapped terms between the source schema terms and the target schema terms and generating rule based matches between source schema terms and target schema terms; based on semantic matching of the source schema terms, retrieving feature store based mappings from the feature store for the unmapped terms between the source schema terms and the target schema terms that were not matched using the rule based mappings, wherein the semantic matching comprises classifying named entities mentioned in unstructured text into pre-defined categories using a trained named entity recognition (NER) model; generating one or more similarity scores for the feature store based mappings of the source schema terms to the target schema terms; and generating possible feature store based mappings for the extracted source schema terms and target schema terms based on the similarity scores; wherein the feature store comprises key-value pairs with keys corresponding to source schema terms and values corresponding to matched target schema terms; transferring the underlying data elements from the first database to the second database and, based on the feature store based mappings, automatically transforming by the feature store the source schema terms to corresponding target schema terms for each of the underlying data elements, the transforming comprising converting source messages of the source schema into target messages of the target schema. ### Claim 9 (depends on 8) The computer readable medium of claim 8, the generating one or more similarity scores comprises using the trained named entity recognition (NER) model. ### Claim 10 (depends on 9) The computer readable medium of claim 9, the mapping further comprising: receiving one or more manual mappings of the source schema terms to the target schema terms in response to the possible feature store based mappings; re-training the trained NER model based on the manual mappings. ### Claim 11 (depends on 10) The computer readable medium of claim 10, further comprising updating the feature store with the re-trained NER model. ### Claim 12 (depends on 8) The computer readable medium of claim 8, wherein the retrieving feature store based mappings from the feature store for the unmapped elements occurs after the retrieving rule based mappings for any remaining unmapped elements. ### Claim 13 (depends on 12) The computer readable medium of claim 12, wherein the retrieving feature store based mappings from the feature store for the unmapped terms occurs after the retrieving rule based mappings for any remaining unmapped terms. ### Claim 14 (depends on 8) The computer readable medium of claim 8, wherein the source schema and target schema both comprise an Extensible Markup Language (XML) schema or a JavaScript Object Notation (JSON) schema and the keys in the feature store are organized based on domain, sub domain, entities and values. ### Claim 15 A cloud infrastructure comprising: a mapping service configured to map a source schema of a first cloud based application to a target schema of a second cloud based application using a cloud based feature store, wherein the feature store comprises a central repository of features used for machine learning model training, the mapping service comprising one or more hardware processors executing instructions, the mapping comprising: receiving a file comprising a plurality of source schema terms of the first application and a plurality of target schema terms of the second application, the file comprising a plurality of unmapped terms, wherein the source schema specifies how each of the source schema terms provide a first application name for each underlying data element of first application data stored in a first database, and the target schema specifies how the each of the target schema terms provide a second application name for each underlying data element of second application data stored in a second database, wherein the underlying data elements stored in the first database are the same as the underlying data elements stored in the second database; parsing the source schema and the target schema and extracting source schema terms and target schema terms that correspond to the underlying data elements; retrieving rule based mappings for the unmapped terms between the source schema terms and the target schema terms and generating rule based matches between source schema terms and target schema terms; based on semantic matching of the source schema terms, retrieving feature store based mappings from the feature store for the unmapped terms between the source schema terms and the target schema terms that were not matched using the rule based mappings, wherein the semantic matching comprises classifying named entities mentioned in unstructured text into pre-defined categories using a trained named entity recognition (NER) model; generating one or more similarity scores for the feature store based mappings of the source schema terms to the target schema terms; and generating possible feature store based mappings for the extracted source schema terms and target schema terms based on the similarity scores; wherein the feature store comprises key-value pairs with keys corresponding to source schema terms and values corresponding to matched target schema terms; transferring the underlying data elements from the first database to the second database and, based on the feature store based mappings, automatically transforming by the feature store the source schema terms to corresponding target schema terms for each of the underlying data elements, the transforming comprising converting source messages of the source schema into target messages of the target schema. ### Claim 16 (depends on 15) The cloud infrastructure of claim 15, the generating one or more similarity scores comprises using the trained named NER model. ### Claim 17 (depends on 16) The cloud infrastructure of claim 16, the mapping further comprising: receiving one or more manual mappings of the source schema terms to the target schema terms in response to the possible feature store based mappings; re-training the trained NER model based on the manual mappings. ### Claim 18 (depends on 17) The cloud infrastructure of claim 17, the mapping further comprising updating the feature store with the re-trained NER model. ### Claim 19 (depends on 15) The cloud infrastructure of claim 15, wherein the retrieving feature store based mappings from the feature store for the unmapped terms occurs after the retrieving rule based mappings for any remaining unmapped terms. ### Claim 20 (depends on 15) The cloud infrastructure of claim 15, wherein the source schema and target schema both comprise an Extensible Markup Language (XML) schema or a JavaScript Object Notation (JSON) schema and the keys in the feature store are organized based on domain, sub domain, entities and values.