Natural language keyword tag extraction

Patent metadata

Field	Value
Publication number	US-12475130-B2
Application number	US-202117544404-A
Country	US
Kind code	B2
Filing date	Dec 7, 2021
Priority date	Aug 23, 2019
Publication date	Nov 18, 2025
Grant date	Nov 18, 2025

Abstract

Various embodiments are generally directed to techniques to provide specific vehicle recommendations to generic user requests. Various techniques, methods, systems, and apparatuses include utilizing one or more tags generated by application of a machine learning model to a data source, where the data source may include generic and specific language with respect to one or more automobiles or vehicles, to provide a recommendation for a particular automobile in response to a user request for a suggestion.

First claim

What is claimed is:

1 . A method, comprising:
receiving a request for an automobile selection; accessing a first plurality of tags related to a first automobile make and model of a plurality of automobile make and models, wherein the first plurality of tags are based on:
a term frequency-inverse document frequency (TF-IDF) model applied to a plurality of terms in a data source comprising a corpus of automobile reviews to compute a respective score for each of the terms, wherein the respective scores of a first subset of the plurality of terms exceed a threshold score; and
an aggregation of the terms of the first subset, wherein:
each tag of the first plurality of tags is associated with a respective plurality of probability distributions generated by the TF-IDF model, wherein each probability distribution associated with one of a plurality of vehicular features, the plurality of vehicular features comprising one of a vehicle component or an operating characteristic of the first automobile make and model, and wherein the first plurality of tags are further based on the respective plurality of probability distributions, and
the aggregation comprises a plurality of documents, each of the plurality of documents corresponding to a different one of the plurality of vehicular features;

returning a first tag of the first plurality of tags as a first suggestion responsive to the request based on:
a relationship between the first tag of the first plurality of tags and a second tag of a second plurality of tags, and
the first tag having a higher score than the second tag; and
updating, by the TF-IDF model based on a selection of the first automobile make and model, the plurality of probability distributions of each tag of the first plurality of tags.

2 . The method of claim 1 , further comprising:
transmitting the first plurality of tags in rank-order based on the scores of the terms as a second suggestion for a particular automobile selection.

3 . The method of claim 2 , wherein the request is a natural language request, wherein the transmitting the first plurality of tags as the second suggestion comprises:
filtering the natural language request using the first plurality of tags, wherein the filtering automatically maps the first plurality of tags to the natural language request.

4 . The method of claim 2 , further comprising:
transmitting a third suggestion specifying to select a second automobile make and model of the plurality of automobile make and models based on the second plurality of tags.

5 . The method of claim 1 , wherein the first plurality of tags and the second plurality of tags are related to one another, and wherein the second plurality of tags are based on a third document of the plurality of documents that includes terms related solely to the second one of the plurality of automobile make and models.

6 . The method of claim 1 , the plurality of vehicular features comprising at least two of engine type, fuel efficiency, engine size, or horsepower.

7 . A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a processor, cause the processor to:
receive a request for an automobile selection; access a first plurality of tags related to a first automobile make and model of a plurality of automobile make and models, wherein the first plurality of tags are based on:
a term frequency-inverse document frequency (TF-IDF) model applied to a plurality of terms in a data source comprising a corpus of automobile reviews to compute a respective score for each of the terms, wherein the respective scores of a first subset of the plurality of terms exceed a threshold score; and
an aggregation of the terms of the first subset, wherein:
each tag of the first plurality of tags is associated with a respective plurality of probability distributions generated by the TF-IDF model, wherein each probability distribution associated with one of a plurality of vehicular features, the plurality of vehicular features comprising one of a vehicle component or an operating characteristic of the first automobile make and model, and wherein the first plurality of tags are further based on the respective plurality of probability distributions, and
the aggregation comprises a plurality of documents, each of the plurality of documents corresponding to a different one of the plurality of vehicular features;

return a first tag of the first plurality of tags as a first suggestion responsive to the request based on:
a relationship between the first tag of the first plurality of tags and a second tag of a second plurality of tags, and
the first tag having a higher score than the second tag; and
update, by the TF-IDF model based on a selection of the first automobile make and model, the plurality of probability distributions of each tag of the first plurality of tags.

8 . The computer-readable storage medium of claim 7 , wherein the instructions further cause the processor to:
transmit the first plurality of tags in rank-order based on the scores of the terms as a second suggestion for a particular automobile selection.

9 . The computer-readable storage medium of claim 8 , wherein the request is a natural language request, wherein the transmitting the first plurality of tags as the second suggestion comprises:
filtering the natural language request using the first plurality of tags, wherein the filtering automatically maps the first plurality of tags to the natural language request.

10 . The computer-readable storage medium of claim 8 , wherein the instructions further cause the processor to:
transmit a third suggestion specifying to select a second automobile make and model of the plurality of automobile make and models based on the second plurality of tags.

11 . The computer-readable storage medium of claim 8 , wherein the first plurality of tags and the second plurality of tags are related to one another, and wherein the second plurality of tags are based on a third document of the plurality of documents that includes terms related solely to the second one of the plurality of automobile make and models.

12 . The computer-readable storage medium of claim 7 , the plurality of vehicular features comprising at least two of engine type, fuel efficiency, engine size, or horsepower.

13 . A computing apparatus comprising:
a processor; and a memory storing instructions that, when executed by the processor, cause the processor to:
receive a request for an automobile selection;
access a first plurality of tags related to a first automobile make and model of a plurality of automobile make and models, wherein the first plurality of tags are based on:
a term frequency-inverse document frequency (TF-IDF) model applied to a plurality of terms in a data source comprising a corpus of automobile reviews to compute a respective score for each of the terms, wherein the respective scores of a first subset of the plurality of terms exceed a threshold score; and
an aggregation of the terms of the first subset, wherein:
each tag of the first plurality of tags is associated with a respective plurality of probability distributions generated by the TF-IDF model, wherein each probability distribution associated with one of a plurality of vehicular features, the plurality of vehicular features comprising one of a vehicle component or an operating characteristic of the first automobile make and model, and wherein the first plurality of tags are further based on the respective plurality of probability distributions, and
the aggregation comprises a plurality of documents, each o

Assignees

• Capital One Services Llc

Inventors

• Price Micah
• Tang Qiaochu
• Hoover Jason
• Dagley Geoffrey
• Ghamsari Avid

Classifications

• G06F16/9535

  Search customisation based on user profiles and personalisation · CPC title

• G06F16/288

  Entity relationship models · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F16/24578 Primary

  using ranking · CPC title

• G06F16/243 Primary

  Natural language query formulation · CPC title