# US 12,182,086 B2 · Generating and executing automatic suggestions to modify data of ingested data collections without additional data ingestion > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,182,086 B2 - Document ID: US12182086B2 - Status: Granted - Kind: B2 - Application number: 17/347,133 - Filed: 2021-06-14 - Published: 2022-12-15 - Granted: 2024-12-31 - Earliest claimed: 2021-06-14 - Assignees: Adobe Inc. - Applicants: Adobe Inc. - Inventors: Ritwik Sinha; Saayan Mitra; Handong Zhao; Somdeb Sarkhel; Trevor Paulsen; William Brandon George - Web page: https://patentgrep.com/patent/US12182086B2 ## CPC - G06F 16/215: Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors - G06N 3/045: Combinations of networks - G06F 16/2423: {Interactive query statement specification based on a database schema} - G06N 3/084: Backpropagation, e.g. using gradient descent - G06N 3/09: Supervised learning - G06N 5/04: Inference or reasoning models - G06N 20/00: Machine learning - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract The present disclosure relates to systems, non-transitory computer-readable media, and methods for generating automatic suggestions to effectively modify the organization of an ingested data collection without destruction of the underlying raw data. In particular, in one or more embodiments, the disclosed systems utilize multiple machine learning models in sequence to determine likelihoods that the organizational structure of an ingested data collection should be modified in various ways. In response to generating these likelihoods, the disclosed systems generate corresponding automatic suggestions to modify the organization of the ingested data collection. In response to a detected selection of one or more of the automatic suggestions, the disclosed systems read data out of the ingested data collection in accordance with the selected automatic suggestions to effectively modify the organization of the ingested data collection. ## Claims ### Claim 1 A computer-implemented method comprising: accessing, by a digital data analytics management system, an ingested data collection categorized into a plurality of dimensions, wherein the ingested data collection includes raw analytics data associated with touchpoints of a plurality of client devices with a third-party computing system ingested into a schema and each of the plurality of dimensions comprises dimension items that indicate pre-defined inputs for interactions between the plurality of client devices and the third-party computing system; generating a dimension report for the plurality of dimensions comprising the dimension items based on the ingested data collection to display to a client device of the digital data analytics management system, the dimension report showing the raw analytics data organized according to the pre-defined inputs for interactions between the plurality of client devices and the third-party computing system; utilizing dimension items from the plurality of dimensions to select one or more machine learning models from a plurality of machine learning models; in response to a query to modify an organization of the ingested data collection, generating one or more automatic suggestions to modify one or more dimensions or dimension items utilizing the selected one or more machine learning models; generating a display of the one or more automatic suggestions for the client device of the digital data analytics management system; and in response to a detected selection of an automatic suggestion from the display of the client device of the digital data analytics management system, modifying the one or more dimensions or dimension items when retrieving data from the ingested data collection by: generating a fallback regular expression incorporating the automatic suggestion; identifying a dimension item or dimension of the one or more dimensions or dimension items referred to by the fallback regular expression; generating a meta-field according to the fallback regular expression for the dimension item or dimension of the one or more dimensions or dimension items; and reading data from the identified dimension item or dimension of the one or more dimensions or dimension items into the meta-field according to instructions within the fallback regular expression without destroying the data ingested into the schema; and generating an updated dimension report for the identified dimension item or dimension of the one or more dimensions or dimension items by reading data out of the meta-field, the meta-field indicating at least one of an indication to disregard, an indication to merge, or an indication to rename as referred to by the fallback regular expression. ### Claim 2 (depends on 1) The computer-implemented method as recited in claim 1, wherein generating the one or more automatic suggestions to modify the one or more dimensions or dimension items comprises: generating, without user input and utilizing a first machine learning model trained for determining a likelihood that dimensions or dimension items in the ingested data collection should be merged an automatic suggestion to merge two or more dimensions or dimension items; generating, without user input and utilizing a second machine learning model trained for determining a likelihood that dimension items in the ingested data collection should be renamed, an automatic suggestion to remove the one or more dimensions or dimension items; or generating, without user input and utilizing a third machine learning model trained for determining a likelihood that dimension items in the ingested data collection should be removed, an automatic suggestion to rename the one or more dimensions or dimension items. ### Claim 3 (depends on 2) The computer-implemented method as recited in claim 2, wherein generating the automatic suggestion to merge the two or more dimensions or dimension items comprises: determining pairwise distances between pairs of names of the one or more dimensions or dimension items; generating a first cluster comprising a first subset of pairs with pairwise distances that are less than a predetermined threshold distance; and generating a first automatic suggestion to merge the one or more dimensions or dimension items within the first cluster. ### Claim 4 (depends on 3) The computer-implemented method as recited in claim 3, wherein generating the first automatic suggestion to merge the two or more dimensions or dimension items within the first cluster further comprises: determining a dimension or dimension item with a corresponding highest number of associated values; and further generating the first automatic suggestion to merge remaining dimension or dimension items within the first cluster with the dimension or dimension item with the corresponding highest number of associated values. ### Claim 5 (depends on 3) The computer-implemented method as recited in claim 3, further comprising: generating a second cluster comprising a second subset of pairs with pairwise distances that are less than the predetermined threshold distance; and generating a second automatic suggestion to merge two or more additional dimensions or dimension items within the second cluster. ### Claim 6 (depends on 5) The computer-implemented method as recited in claim 5, further comprising: determining inter-cluster distances for the first cluster and the second cluster; and ranking the first automatic suggestion and the second automatic suggestion based on the inter-cluster distances. ### Claim 7 (depends on 2) The computer-implemented method as recited in claim 2, wherein generating the automatic suggestion to remove the one or more dimensions or dimension items comprises: generating embeddings for the one or more dimensions or dimension items within a common space; determining an average similarity score for each embedding, wherein the average similarity score represents an average distance between each embedding and other embeddings within the common space; determining one or more embeddings with average similarity scores that do not satisfy a predetermined cut-off similarity score; and generating the automatic suggestion to remove the one or more dimensions or dimension items corresponding to the determined one or more embeddings with the average similarity scores that do not satisfy the predetermined cut-off similarity score. ### Claim 8 (depends on 2) The computer-implemented method as recited in claim 2, wherein generating the automatic suggestion to rename the one or more dimensions or dimension items comprises: determining names for the one or more dimensions or dimension items; generating embeddings of the names within a common space; and for each embedding of the embeddings, iteratively: mask a portion of each embedding; predict a string corresponding to the masked portion of each embedding; compare the predicted string to the portion of a name corresponding to the masked portion of each embedding; and generating the one or more automatic suggestions to rename the one or more dimensions or dimension items based on the comparison. ### Claim 9 (depends on 1) The computer-implemented method as recited in claim 1, further comprising: in response to the query to modify the organization of the ingested data collection, generating, the one or more automatic suggestions to modify two or more dimensions or dimension items; and generating a first meta-field and a second meta-field for the two or more dimensions or dimension items, wherein the first meta-field differs from the second meta-field. ### Claim 10 (depends on 9) The computer-implemented method as recited in claim 9, further comprising: modifying the organization of the ingested data collection by reading data from the ingested data collection according to the first meta-field and the second meta-field, wherein modifying the organization of the ingested data collection occurs from changing how data for the two or more dimensions or dimension items are read from the ingested data collection based on the first meta-field and the second meta-field. ### Claim 11 A non-transitory computer-readable storage medium comprising instructions that, when executed by at least one processor, cause a computing device to: access, by a digital data analytics management system, an ingested data collection categorized into a plurality of dimensions, wherein the ingested data collection includes raw analytics data associated with touchpoints of a plurality of client devices with a third-party computing system ingested into a schema and each of the plurality of dimensions comprises dimension items that indicate pre-defined inputs for interactions between the plurality of client devices and the third-party computing system; generate a dimension report for the plurality of dimensions comprising the dimension items based on the ingested data collection to display to a client device of the digital data analytics management system, the dimension report showing the raw analytics data organized according to the pre-defined inputs for interactions between the plurality of client devices and the third-party computing system; in response to a query comprising a detected selection in connection with a dimension from the plurality of dimensions: utilize the dimension from the plurality of dimensions to select one or more machine learning models from a plurality of machine learning models; generate one or more automatic suggestions to modify one or more dimension items of the dimension utilizing the selected one or more machine learning models; generate a display of the one or more automatic suggestions for the client device of the digital data analytics management system; and in response to a detected selection of an automatic suggestion from the display of the client device of the digital data analytics management system: determine one or more dimension items of the dimension that correspond to the automatic suggestion, generate a fallback regular expression incorporating the automatic suggestion, identify a dimension item of the one or more dimension items referred to by the fallback regular expression, generate a meta-field according to the fallback regular expression for the dimension item of the one or more dimension items, and read data from the identified dimension item of the one or more dimension items into the meta-field according to instructions within the fallback regular expression without destroying the data ingested into the schema; and generate an updated dimension report for the identified dimension item or dimension of the one or more dimensions or dimension items by reading data out of the meta-field, the meta-field indicating at least one of an indication to disregard, an indication to merge, or an indication to rename as referred to by the fallback regular expression. ### Claim 12 (depends on 11) The non-transitory computer-readable storage medium as recited in claim 11, further comprising instructions that, when executed by the at least one processor, cause the computing device to determine the one or more dimension items of the dimension that correspond to the automatic suggestion by determining two or more dimension items indicated by the automatic suggestion to merge together. ### Claim 13 (depends on 12) The non-transitory computer-readable storage medium as recited in claim 12, further comprising instructions that, when executed by the at least one processor, cause the computing device to modify the one or more dimension items of the dimension when retrieving data from the ingested data collection by: determining a main dimension item of the two or more dimension items; and renaming remaining dimension items of the two or more dimension items by reading the remaining dimension items into the meta-field named to mirror a name of the main dimension item. ### Claim 14 (depends on 11) The non-transitory computer-readable storage medium as recited in claim 11, further comprising instructions that, when executed by the at least one processor, cause the computing device to: determine the one or more dimension items of the dimension associated with the automatic suggestion by determining a dimension item with a name corresponding with a dimension item indicated by the automatic suggestion to remove; and modify the one or more dimension items of the dimension when retrieving data from the ingested data collection by removing the determined dimension from the dimension when reading data from the ingested data collection in response to a query. ### Claim 15 (depends on 11) The non-transitory computer-readable storage medium as recited in claim 11, further comprising instructions that, when executed by the at least one processor, cause the computing device to access the ingested data collection categorized into the plurality of dimensions, wherein the plurality of dimensions comprises geographic location, browser, referrer, search engine, device type, product, webpage, gender, purchase, downloads, age, and digital content campaign. ### Claim 16 (depends on 11) The non-transitory computer-readable storage medium as recited in claim 11, further comprising instructions that, when executed by the at least one processor, cause the computing device to determine the one or more dimension items of the dimension associated with the automatic suggestion by determining a dimension item with a name corresponding with a dimension item indicated by the automatic suggestion to rename. ### Claim 17 (depends on 16) The non-transitory computer-readable storage medium as recited in claim 16, further comprising instructions that, when executed by the at least one processor, cause the computing device to: modify the one or more dimension items of the dimension when retrieving data from the ingested data collection by reading the dimension item into the meta-field named to match a name indicated by the automatic suggestion, wherein the meta-field is a storage location separate from the ingested data collection; and read data out of the meta-field in the storage location separate from the ingested data collection to generate the updated dimension report for the identified dimension item or dimension of the one or more dimensions or dimension items. ### Claim 18 A system comprising: at least one computer memory device comprising: a digital data analytics management system; at least one ingested data collection organized into a plurality of dimensions and corresponding dimension items, wherein the at least one ingested data collection includes raw analytics data associated with touchpoints of a plurality of client devices with a third-party computing system ingested into a schema and each of the plurality of dimensions comprises dimension items that indicate pre-defined inputs for interactions between the plurality of client devices and the third-party computing system, and one or more machine learning models for generating automatic modification suggestions; and one or more servers configured to cause the system to: sequentially utilize a plurality of machine learning models to generate one or more automatic suggestions to modify one or more dimension items; generate a dimension report for the plurality of dimensions comprising the dimension items based on the ingested data collection to display to a client device of the digital data analytics management system, the dimension report showing the raw analytics data organized according to the pre-defined inputs for interactions between the plurality of client devices and the third-party computing system; in response to a query to modify an organization of the at least one ingested data collection, generate a display of the one or more automatic suggestions in connection with the dimension report of the one or more dimension items of the plurality of dimensions based on the ingested data collection to provide for display to the client device of the digital data analytics management system; and in response to a detected selection of an automatic suggestion from the display of the client device of the digital data analytics management system: generate a fallback regular expression incorporating the automatic suggestion; identify a dimension item of the one or more or dimension items referred to by the fallback regular expression; generate a meta-field according to the fallback regular expression for the dimension item of the one or more dimension items; and read data from the identified dimension item or dimension of the one or more dimensions or dimension items into the meta-field according to instructions within the fallback regular expression without destroying the data ingested into the schema; and update the dimension report of the one or more dimension items by reading data out of the meta-field, the meta-field indicating at least one of an indication to disregard, an indication to merge, or an indication to rename as referred to by the fallback regular expression to reflect modifications from reading data out of the one or more dimension items into the meta-field according to the fallback regular expression. ### Claim 19 (depends on 18) The system as recited in claim 18, wherein the one or more servers are further configured to cause the system to sequentially utilize the plurality of machine learning models by: utilizing a merge machine learning model to determine a likelihood that two or more dimension items should be merged into one dimension item; upon determining a likelihood that two or more dimension items can be merged, utilizing a renaming machine learning model to determine a likelihood that one or more dimension items are associated with an incorrect word or phrase; and upon determining a likelihood that one or more dimension items are associated with an incorrect word or phrase, utilizing a removal machine learning model to determine a likelihood that one or more dimension items should be removed from a dimension. ### Claim 20 (depends on 18) The system as recited in claim 18, wherein the one or more servers are further configured to cause the system to read the data out of the one or more dimension items according to the fallback regular expression without destroying the data ingested into the schema by: in response to a query that involves the one or more dimension items, utilizing the fallback regular expression as an instruction layer to read data from the at least one ingested data collection; and correcting the one or more dimension items from the query when retrieving data from the at least one ingested data collection according to the instruction layer.