# US 2025/0094504 A1 · Generating Enhanced Queries Using Machine Learning Models

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0094504 A1
- Document ID: US20250094504A1
- Status: Published application
- Kind: A1
- Application number: 18/467,360
- Filed: 2023-09-14
- Published: 2025-03-20
- Earliest claimed: 2023-09-14
- Assignees: Oracle International Corporation
- Applicants: Oracle International Corporation
- Inventors: Venkatakrishnan Gopalakrishnan; May Bich Nhi Lam; Diego Ceferino Torres Dho; Jan Sterba
- Web page: https://patentgrep.com/patent/US20250094504A1

## CPC

- G06F 16/9532: Query formulation
- G06F 16/3347: {using vector based model}
- G06F 16/93: Document management systems
- 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

Techniques for generating terms to replace an initial set of search terms for a query are disclosed. A system generates a training data set for training a machine learning model. Generating the training data set includes generating search value vectors for each of a set of labels based on sets of search values associated respectively with the labels in the set of labels. The system trains a machine learning model to predict a target label for a target search vector based on the set of labels and the respectively associated search value vectors. The system generates a target search value vector based on an initial set of search values. The system then applies the trained machine learning model to the target search value vector to predict the target label. The target label is used as a search term, that replaces the initial set of search values, for executing the query.

## Claims

### Claim 1

A non-transitory computer readable medium comprising instructions which, when executed by one or more hardware processors, causes performance of operations comprising:
obtaining (a) a plurality of labels and (b) a plurality of sets of search values, each of the plurality of labels being associated with a respective set of search values from the plurality of sets of search values;
aggregating the search values in the plurality of the sets of search values to generate a list of n search values;
generating a plurality of search value vectors for the plurality of labels, respectively, at least by:
for each particular label of the plurality of labels, generating a particular search value vector representing information for one or more particular search values in a particular set of search values associated with the particular label, wherein the information for the one or more particular search values comprises one or more of:
(a) locations of the one or more particular search values within the list of n search values,
(b) a term frequency (tf) of the one or more particular search values in the particular set of search values, or
(c) an inverse document frequency (idf) of the one or more particular search values in the list of n search values;
generating training data, for training a machine learning model, based on the plurality of labels and the plurality of search value vectors;
training the machine learning model to predict labels from vectors using the training data;
receiving a query comprising a target set of one or more target search values;
computing a target search value vector representing information for the target set of one or more target search values, wherein the information for the one or more target search values comprises one or more of:
(a) locations of the one or more target search values within the list of n search values,
(b) a term frequency (tf) of the one or more target search values in the target set of search values, or
(c) an inverse document frequency (idf) of the one or more target search values in the list of n search values;
applying the machine learning model to the target search value vector to predict a target label for the query; and
using the target label, as a search term, to execute the query.

### Claim 2 (depends on 1)

The non-transitory computer readable medium of claim 1, wherein the particular search value vector has n vector values that are computed based at least in part on the list of n search values, wherein each vector value x, of the n vector values, is determined based on whether or not any of particular set of search values match the x-th value within the list of n values.

### Claim 3 (depends on 1)

The non-transitory computer readable medium of claim 1, wherein the operations further comprise:
receiving feedback associated with a prediction by the machine learning model; and retraining the machine learning model based on the feedback.

### Claim 4 (depends on 1)

The non-transitory computer readable medium of claim 1, wherein the operations further comprise:
receiving a second query comprising a second target set of one or more target search values;
computing a second target search value vector representing information for the one or more target search values of the second target set, wherein the information for the one or more target search values of the second target set comprises one or more of:
(a) locations of the one or more target search values of the second target set within the list of n search values,
(b) a term frequency (tf) of the one or more target search values of the second target set in the target set of search values of the second target set, or
(c) an inverse document frequency (idf) of the one or more target search values of the second target set in the list of n search values;
applying the machine learning model to the second target search value vector to predict a target label for the second query, wherein the machine learning model returns no target label for the second query; and
using the one or more target search values of the second target set as a search term to execute the second query.

### Claim 5 (depends on 1)

The non-transitory computer readable medium of claim 1, wherein the operations further comprise:
preprocessing the search values, wherein preprocessing the search values comprises one or more of:
(a) removing stop words,
(b) converting search values to lowercase,
(c) cleaning up special characters,
(d) deleting duplicate words within each set of search values,
(e) lemmatizing the search values, or
(f) stemming the search values.

### Claim 6 (depends on 1)

The non-transitory computer readable medium of claim 1, wherein obtaining the plurality of labels and the plurality of sets of search values includes receiving a user prepared document, the user prepared document including the plurality of labels and the plurality of sets of search values.

### Claim 7 (depends on 1)

The non-transitory computer readable medium of claim 1, wherein obtaining the plurality of labels and the plurality of search values further includes obtaining a landing page ID for each of the plurality of labels, wherein applying the machine learning model to the target search value vector to predict a target label for the query further includes providing a landing page ID for the target label.

### Claim 8 (depends on 1)

The non-transitory computer readable medium of claim 1, wherein applying the machine learning model to the target search value vector to predict a target label for the query comprises:
using a K-Nearest Neighbor (KNN) model to determine a closest match to the target search value vector.

### Claim 9

A method comprising:
obtaining (a) a plurality of labels and (b) a plurality of sets of search values, each of the plurality of labels being associated with a respective set of search values from the plurality of sets of search values;
aggregating the search values in the plurality of the sets of search values to generate a list of n search values;
generating a plurality of search value vectors for the plurality of labels, respectively, at least by:
for each particular label of the plurality of labels, generating a particular search value vector representing information for one or more particular search values in a particular set of search values associated with the particular label, wherein the information for the one or more particular search values comprises one or more of:
(a) locations of the one or more particular search values within the list of n search values,
(b) a term frequency (tf) of the one or more particular search values in the particular set of search values, or
(c) an inverse document frequency (idf) of the one or more particular search values in the list of n search values;
generating training data, for training a machine learning model, based on the plurality of labels and the plurality of search value vectors;
training the machine learning model to predict labels from vectors using the training data;
receiving a query comprising a target set of one or more target search values;
computing a target search value vector representing information for the target set of one or more target search values, wherein the information for the one or more target search values comprises one or more of:
(a) locations of the one or more target search values within the list of n search values,
(b) a term frequency (tf) of the one or more target search values in the target set of search values, or
(c) an inverse document frequency (idf) of the one or more target search values in the list of n search values;
applying the machine learning model to the target search value vector to predict a target label for the query; and
using the target label, as a search term, to execute the query.

### Claim 10 (depends on 9)

The method of claim 9, wherein the particular search value vector has n vector values that are computed based at least in part on the list of n search values, wherein each vector value x, of the n vector values, is determined based on whether or not any of particular set of search values match the x-th value within the list of n values.

### Claim 11 (depends on 9)

The method of claim 9, further comprising:
receiving feedback associated with a prediction by the machine learning model; and retraining the machine learning model based on the feedback.

### Claim 12 (depends on 9)

The method of claim 9, further comprising:
receiving a second query comprising a second target set of one or more target search values;
computing a second target search value vector representing information for the one or more target search values of the second target set, wherein the information for the one or more target search values of the second target set comprises one or more of:
(a) locations of the one or more target search values of the second target set within the list of n search values,
(b) a term frequency (tf) of the one or more target search values of the second target set in the target set of search values of the second target set, or
(c) an inverse document frequency (idf) of the one or more target search values of the second target set in the list of n search values;
applying the machine learning model to the second target search value vector to predict a target label for the second query, wherein the machine learning model returns no target label for the second query; and
using the one or more target search values of the second target set as a search term to execute the second query.

### Claim 13 (depends on 9)

The method of claim 9, further comprising:
preprocessing the search values, wherein preprocessing the search values comprises one or more of:
(a) removing stop words,
(b) converting search values to lowercase,
(c) cleaning up special characters,
(d) deleting duplicate words within each set of search values,
(e) lemmatizing the search values, or
(f) stemming the search values.

### Claim 14 (depends on 9)

The method of claim 9, wherein obtaining the plurality of labels and the plurality of sets of search values includes receiving a user prepared document, the user prepared document including the plurality of labels and the plurality of sets of search values.

### Claim 15 (depends on 9)

The method of claim 9, wherein obtaining the plurality of labels and the plurality of search values further includes obtaining a landing page ID for each of the plurality of labels, wherein applying the machine learning model to the target search value vector to predict a target label for the query further includes providing a landing page ID for the target label.

### Claim 16 (depends on 9)

The method of claim 9, wherein applying the machine learning model to the target search value vector to predict a target label for the query comprises:
using a K-Nearest Neighbor (KNN) model to determine a closest match to the target search value vector.

### Claim 17

A system comprising:
one or more processors; and
memory storing instructions that, when executed by the one or more processors, cause the system to perform:
obtaining (a) a plurality of labels and (b) a plurality of sets of search values, each of the plurality of labels being associated with a respective set of search values from the plurality of sets of search values;
aggregating the search values in the plurality of the sets of search values to generate a list of n search values;
generating a plurality of search value vectors for the plurality of labels, respectively, at least by:
for each particular label of the plurality of labels, generating a particular search value vector representing information for one or more particular search values in a particular set of search values associated with the particular label, wherein the information for the one or more particular search values comprises one or more of:
(a) locations of the one or more particular search values within the list of n search values,
(b) a term frequency (tf) of the one or more particular search values in the particular set of search values, or
(c) an inverse document frequency (idf) of the one or more particular search values in the list of n search values;
generating training data, for training a machine learning model, based on the plurality of labels and the plurality of search value vectors;
training the machine learning model to predict labels from vectors using the training data;
receiving a query comprising a target set of one or more target search values;
computing a target search value vector representing information for the target set of one or more target search values, wherein the information for the one or more target search values comprises one or more of:
(a) locations of the one or more target search values within the list of n search values,
(b) a term frequency (tf) of the one or more target search values in the target set of search values, or
(c) an inverse document frequency (idf) of the one or more target search values in the list of n search values;
applying the machine learning model to the target search value vector to predict a target label for the query; and
using the target label, as a search term, to execute the query.

### Claim 18 (depends on 17)

The system of claim 17, wherein the particular search value vector has n vector values that are computed based at least in part on the list of n search values, wherein each vector value x, of the n vector values, is determined based on whether or not any of particular set of search values match the x-th value within the list of n values.

### Claim 19 (depends on 17)

The system of claim 17, wherein the instructions cause the system to further perform:
receiving feedback associated with a prediction by the machine learning model; and retraining the machine learning model based on the feedback.

### Claim 20 (depends on 17)

The system of claim 17, wherein the instructions cause the system to further perform:
receiving a second query comprising a second target set of one or more target search values;
computing a second target search value vector representing information for the one or more target search values of the second target set, wherein the information for the one or more target search values of the second target set comprises one or more of:
(a) locations of the one or more target search values of the second target set within the list of n search values,
(b) a term frequency (tf) of the one or more target search values of the second target set in the target set of search values of the second target set, or
(c) an inverse document frequency (idf) of the one or more target search values of the second target set in the list of n search values;
applying the machine learning model to the second target search value vector to predict a target label for the second query, wherein the machine learning model returns no target label for the second query; and
using the one or more target search values of the second target set as a search term to execute the second query.

### Claim 21 (depends on 17)

The system of claim 17, wherein the instructions cause the system to further perform:
preprocessing the search values, wherein preprocessing the search values comprises one or more of:
(a) removing stop words,
(b) converting search values to lowercase,
(c) cleaning up special characters,
(d) deleting duplicate words within each set of search values,
(e) lemmatizing the search values, or
(f) stemming the search values.

### Claim 22 (depends on 17)

The system of claim 17, wherein obtaining the plurality of labels and the plurality of sets of search values includes receiving a user prepared document, the user prepared document including the plurality of labels and the plurality of sets of search values.

### Claim 23 (depends on 17)

The system of claim 17, wherein obtaining the plurality of labels and the plurality of search values further includes obtaining a landing page ID for each of the plurality of labels, wherein applying the machine learning model to the target search value vector to predict a target label for the query further includes providing a landing page ID for the target label.

### Claim 24 (depends on 17)

The system of claim 17, wherein applying the machine learning model to the target search value vector to predict a target label for the query comprises:
using a K-Nearest Neighbor (KNN) model to determine a closest match to the target search value vector.
