# US 2022/0035722 A1 · Sparse Intent Clustering Through Deep Context Encoders

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

- Number: US 2022/0035722 A1
- Document ID: US20220035722A1
- Status: Published application
- Kind: A1
- Application number: 16/983,146
- Filed: 2020-08-03
- Published: 2022-02-03
- Earliest claimed: 2020-08-03
- Assignees: ADP, Inc.
- Applicants: Adp, LLC
- Inventors: Allan Barcelos; Fernanda Tosca; Israel Oliveira; Leandro Bianchini; Renata Palazzo
- Web page: https://patentgrep.com/patent/US20220035722A1

## CPC

- G06F 11/3072: {where the reporting involves data filtering, e.g. pattern matching, time or event triggered, adaptive or policy-based reporting}
- G06F 40/30: Semantic analysis
- G06F 16/35: Clustering; Classification
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0495: Quantised networks; Sparse networks; Compressed networks
- G06F 11/3438: {monitoring of user actions (tracking the activity of the user H04L67/535)}
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method of sparse intent clustering is provided. The method comprises identifying features in a number of electronic user reports created by a user and contained in a database, wherein the features include a title and description. The features of each user report are encoded into a binary vector. The binary vector for each user report is fed into an autoencoder neural network that creates a N-dimensional vector representing the user report. The float vectors representing the user reports are projected into a N-dimensional space. The float vectors are clustered according to cosine similarities, wherein each vector cluster represents an intent of the user in creating the reports. The intent of each vector cluster is then labeled.

## Claims

### Claim 1

A computer-implemented method of sparse intent clustering, the method comprising:
using a number of processors to perform the steps of:
identifying features in a number of electronic user reports created by a user and contained in a database, wherein the features include a title and description;
encoding the features of each user report into a binary vector;
feeding the binary vector for each user report into an autoencoder neural network, wherein the autoencoder neural network generates a N-dimensional float vector representing the user report;
projecting the float vectors representing the user reports into a N-dimensional space;
clustering the float vectors into vector clusters according to cosine similarities, wherein each vector cluster represents an intent of the user in creating the reports; and
labeling the intent of each vector cluster.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
receiving input from the user to create a new user report, wherein the input includes a title and description for the new user report;
determining an intent of the new user report according to report data;
identifying a vector cluster labeled with an intent that matches the intent of the new user report; and
providing the user with suggested features for the new user report according to features in the vector cluster identified with the matching intent.

### Claim 3 (depends on 1)

The method of claim 1, wherein encoding the features of each user report comprises one-hot encoding a number of fields in the user report.

### Claim 4 (depends on 1)

The method of claim 1, wherein encoding the features of each user report comprises:
decomposing equations of a number of filters in the user report to produce decomposed filters; and
one-hot encoding the decomposed filters.

### Claim 5 (depends on 1)

The method of claim 1, wherein encoding the features of each user report comprises:
decomposing equations of a number of custom fields in the user report to produce decomposed custom fields; and
one-hot encoding the decomposed custom fields.

### Claim 6 (depends on 1)

The method of claim 1, wherein encoding the features of each user report comprises encoding the title and description with embeddings from language models.

### Claim 7 (depends on 1)

The method of claim 1, wherein vectors within each cluster have cosine similarities above 0.5.

### Claim 8

A system for sparse intent clustering, the system comprising:
a storage device configured to store program instructions; and
one or more processors operably connected to the storage device and configured to execute the program instructions to cause the system to:
identify features in a number of electronic user reports created by a user and contained in a database, wherein the features include a title and description;
encode the features of each user report into a binary vector;
feed the binary vector for each user report into an autoencoder neural network, wherein the autoencoder neural network generates a N-dimensional float vector representing the user report;
project the float vectors representing the user reports into a N-dimensional space;
cluster the float vectors into vector clusters according to cosine similarities, wherein each vector cluster represents an intent of the user in creating the reports; and
label the intent of each vector cluster.

### Claim 9 (depends on 8)

The system of claim 8, wherein the processors further execute instructions to:
receive input from the user to create a new user report, wherein the input includes a title and description for the new user report;
determine an intent of the new user report according to report data;
identify a vector cluster labeled with an intent that matches the intent of the new user report; and
provide the user with suggested features for the new user report according to features in the vector cluster identified with the matching intent.

### Claim 10 (depends on 8)

The system of claim 8, wherein encoding the features of each user report comprises one-hot encoding a number of fields in the user report.

### Claim 11 (depends on 8)

The system of claim 8, wherein encoding the features of each user report comprises:
decomposing equations of a number of filters in the user report to produce decomposed filters; and
one-hot encoding the decomposed filters.

### Claim 12 (depends on 8)

The system of claim 8, wherein encoding the features of each user report comprises:
decomposing equations of a number of custom fields in the user report to produce decomposed custom fields; and
one-hot encoding the decomposed custom fields.

### Claim 13 (depends on 8)

The system of claim 8, wherein encoding the features of each user report comprises encoding the title and description with embeddings from language models.

### Claim 14

A computer program product for sparse intent clustering, the computer program product comprising:
a computer-readable storage medium having program instructions embodied thereon to perform the steps of:
identifying features in a number of electronic user reports created by a user and contained in a database, wherein the features include a title and description;
encoding the features of each user report into a binary vector;
feeding the binary vector for each user report into an autoencoder neural network, wherein the autoencoder neural network generates a N-dimensional float vector representing the user report;
projecting the float vectors representing the user reports into a N-dimensional space;
clustering the float vectors into vector clusters according to cosine similarities, wherein each vector cluster represents an intent of the user in creating the reports; and
labeling the intent of each vector cluster.

### Claim 15 (depends on 14)

The computer program product of claim 14, further comprising instructions for:
receiving input from the user to create a new user report, wherein the input includes a title and description for the new user report;
determining an intent of the new user report according to report data;
identifying a vector cluster labeled with an intent that matches the intent of the new user report; and
providing the user with suggested features for the new user report according to features in the vector cluster identified with the matching intent.

### Claim 16 (depends on 14)

The computer program product of claim 14, wherein encoding the features of each user report comprises one-hot encoding a number of fields in the user report.

### Claim 17 (depends on 14)

The computer program product of claim 14, wherein encoding the features of each user report comprises:
decomposing equations of a number of filters in the user report to produce decomposed filters; and
one-hot encoding the decomposed filters.

### Claim 18 (depends on 14)

The computer program product of claim 14, wherein encoding the features of each user report comprises:
decomposing equations of a number of custom fields in the user report to produce decomposed custom fields; and
one-hot encoding the decomposed custom fields.

### Claim 19 (depends on 14)

The computer program product of claim 14, wherein encoding the features of each user report comprises encoding the title and description with embeddings from language models.

### Claim 20 (depends on 14)

The computer program product of claim 14, wherein vectors within each cluster have cosine similarities above 0.5.