Techniques and systems for smart natural language processing of search results

Patent metadata

Field	Value
Publication number	US-11934413-B2
Application number	US-202117530149-A
Country	US
Kind code	B2
Filing date	Nov 18, 2021
Priority date	Nov 18, 2021
Publication date	Mar 19, 2024
Grant date	Mar 19, 2024

Abstract

A system, processes, and a computer-readable storage medium are provided method includes receiving a search query including one or more of a textual phrase, a document, or data including metadata from a user. For example, a processor may execute a theme-generating machine learning algorithm to generate a theme of the search query. Profiles of other users are identified as similar to a profile of the user may be located. A search history of each of the located other users may be located in the collected search history. Content in the searchable content that corresponds to the generated theme of the search query and according to a result of the evaluation of the located search history of each of the other users may be obtained. The obtained content may be scored based on a scoring algorithm and output a set of scored search results to the input/output device.

First claim

What is claimed is:

1. A system, comprising:
a memory storing programming code including a search result scoring application, a plurality of user profiles, and collected search history, wherein the collected search history is a database that includes search terms used in a query, results from the query, and a degree of engagement with results of interest of a plurality of users;
an input/output device; and
a processor operable to:
receive, via the input/output device, a search query including one or more of a textual phrase, a document, or data including metadata from a user;
generate a theme of the search query based on an evaluation of the search query;
locate other users having profiles similar to a user profile of the user from whom the search query was received, wherein the user profile is based on a user group to which the user and the other users are assigned;
obtain a plurality of search histories by retrieving a search history of each of the located other users from the collected search history;
identify content corresponding to the generated theme in at least one of the plurality of search histories;
execute the search query to obtain content corresponding to the generated theme of the search query from a searchable content module, wherein the searchable content module includes public source content and internal source content, the execution of the search query to identify the public source content from a public source having a similar theme as the search query and identify the internal source content from an internal source searchable content repository based on a work assignment category obtained from the user profile;
score the obtained content and the identified content using a scoring algorithm based on a result of an evaluation of search history results considered relevant to each user for relevancy to the generated theme of the search query, wherein the score is based on application of the search result scoring application to the identified public source content and the identified internal source content; and
output a set of scored search results to the input/output device.

2. The system of claim 1 , wherein the processor, when generating the generated theme of the search query, is further operable to:
evaluate the search query using a natural language processing algorithm;
evaluate the metadata related to the textual phrase, the document or the data;
input results of the evaluation of the metadata ad the results of the natural language processing algorithm into a theme-generating machine learning algorithm; and
receive an indication of a theme of the search query based on the results of the theme-generating machine learning algorithm.

3. The system of claim 1 , wherein the processor, when locating the other users, is further operable to:
identify the other users in the plurality of user profiles having a work assignment category that is a same work assignment category as in the user profile; and
group the identified other users in the user group with the user.

4. The system of claim 3 , wherein the processor, when locating the other users, is further operable to:
identify additional users, not yet included in the user group, that have a work assignment category related to the work assignment category in the user profile, wherein users with the related work assignment category perform data system searches using a similar search query that has search terms similar to the received search query; and
group the identified additional users in the user group with the identified other users.

5. The system of claim 1 , wherein the processor, when obtaining a plurality of search histories by retrieving a search history of each of the located other users, is further operable to:
identify past search queries of each user having a profile similar to the user that have a theme that is substantially similar to the generated theme; and
retrieve content from the identified past search queries that was indicated as relevant by a respective user of the identified past search queries.

6. The system of claim 1 , wherein the processor, when obtaining the content corresponding to the generated theme, is further operable to:
identify partner source content from a partner source search content repository of searchable content; and
input the identified partner source content with the identified public source content and identified internal source content into the search result scoring application, wherein the search result scoring application is further applied to the identified partner source content.

7. The system of claim 1 , wherein the processor, when scoring the obtained content and the identified content, is further operable to:
apply a past preference benchmark to the search query in a machine learning algorithm;
determine a past preference score of the search query from the machine learning algorithm based on the applied past preference benchmark;
apply a work assignment category benchmark to the search query in the scoring algorithm;
determine a work assignment category score of the search query from the scoring algorithm;
provide the past preference score and the work assignment category score to the search result scoring application; and
generate a relevancy score for the obtained content based a result of the search result scoring application using the past preference score, and the work assignment category score.

8. A method, comprising:
receiving, via an input/output device, a search query including one or more of a textual phrase, a document, or data including metadata;
generating, by a processor executing a theme-generating machine learning algorithm, a theme of the search query;
locating profiles of other users from a plurality of user profiles that are identified as similar to a user profile of a user from whom the search query was received based on a group to which the other users are assigned;
locating a search history of each of the other users having a profile identified as similar to the user profile in a collected search history;
evaluating, by the processor, the located search history of each of the other users having the profile identified as similar to the user profile and the user in the located search history based on the generated theme of the search query;
obtaining content in a searchable content that corresponds to the generated theme of the search query and according to a result of evaluation of the located search history of each of the other users, wherein the searchable content includes public source content and internal source content, the obtaining content to identify the public source content from a public source having a similar theme as the search query and to identify the internal source content from an internal source searchable content repository of the searchable content based on a work assignment category obtained from the user profile;
scoring the obtained content based on a scoring algorithm, the score based on application of a search result scoring application to the identified public source content and the identified internal source content; and
outputting a set of scored search results to the input/output device.

9. The method of claim 8 , wherein when generating the theme of the search query, the method further comprises:
evaluating the search query using a natural language processing algorithm;
inputting results of the natural language processing algorithm into the theme-generating machine learning algorithm;
evaluating the metadata related to the textual phrase, the document or the data;
inputting results of the evaluation of the metadata into the theme-generating machine learning algorithm; and
based on an output of the theme-generati

Assignees

• Capital One Services Llc

Inventors

• Coble Piper Alexandra
• Kuhn Natalie
• Cho Paul
• Parker Ryan M
• Petrunin Sergey
• Williams Gary B
• Campbell Eric
• Cheng Lin Ni Lisa
• Sullivan Matthew Kevin
• Demchalk Chris

Classifications

• G06F16/24578 Primary

  using ranking · CPC title

• G06F16/243 Primary

  Natural language query formulation · CPC title

• G06F16/285

  Clustering or classification · CPC title