# US 2021/0343406 A1 · Integrative machine learning framework for combining sentiment-based and symptom-based predictive inferences

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

- Number: US 2021/0343406 A1
- Document ID: US20210343406A1
- Status: Published application
- Kind: A1
- Application number: 16/862,943
- Filed: 2020-04-30
- Published: 2021-11-04
- Earliest claimed: 2020-04-30
- Assignees: Optum, Inc.
- Applicants: Optum, Inc.
- Inventors: Corey D. McMillan; William Alexander Wood; Austin Schick; Donald Landrum; Columbus Dong; Rachel Szabo; William Drayton Taylor; Austin T. Gansz
- Web page: https://patentgrep.com/patent/US20210343406A1

## CPC

- G06F 40/30: Semantic analysis
- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G06F 16/244: {Grouping and aggregation}
- G06F 16/345: {Summarisation for human users}
- G06F 40/289: Phrasal analysis, e.g. finite state techniques or chunking
- G06N 3/045: Combinations of networks
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 3/09: Supervised learning
- G06N 3/096: Transfer learning
- G06N 20/00: Machine learning
- G16H 10/60: for patient-specific data, e.g. for electronic patient records
- G16H 10/20: for electronic clinical trials or questionnaires
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Techniques for integrative machine learning using sentiment-based predictive inferences and symptom-based predictive are discussed herein. In one example, a method includes determining, based on one or more health monitoring logs, a first distribution of symptomatic prediction labels over a first period of time associated with the one or more health monitoring logs; processing the one or more health monitoring logs and using a sentiment detection machine learning model to determine a second distribution of extracted sentiment scores over the first period of time; generating, based on the first distribution and the second distribution, an aggregate distribution of inferred health-related predictions over the first period of time; and causing display of an aggregate distribution user interface that is configured to display the aggregate distribution.

## Claims

### Claim 1

A computer-implemented method for generating one or more health-related predictive inferences from one or more health monitoring logs, the computer-implemented method comprising:
determining, by one or more processors and based on the one or more health monitoring logs, a first distribution of symptomatic prediction labels over a first period of time associated with the one or more health monitoring logs;
processing by the one or more processors, the one or more health monitoring logs and using a sentiment detection machine learning model to determine a second distribution of extracted sentiment scores over the first period of time;
generating, by the one or more processors and based on the first distribution and the second distribution, an aggregate distribution of inferred health-related predictions over the first period of time; and
providing, by the one or more processors, a presentation for display via a user interface, wherein the presentation comprises the aggregate distribution.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein determining the first distribution comprises:
determining, for each health monitoring log of the one or more health monitoring logs, one or more extracted symptoms;
identifying, for each time unit of one or more time units associated with the first period of time, one or more first health monitoring logs of the one or more health monitoring logs associated with the time unit;
determining, for each time unit of the one or more time units, one or more first extracted symptoms, wherein the one or more first extracted symptoms for a time unit of the one or more time units include each one or more extracted symptoms for a first health monitoring log of the one or more first health monitoring logs that is associated with the time unit; and
combining each one or more first extracted symptoms associated with a time unit of the one or more time units to determine the first distribution.

### Claim 3 (depends on 2)

The computer-implemented method of claim 2, wherein determining the one or more extracted symptoms for a first health monitoring log of the one or more journal entries comprises:
identifying one or more relevant phrases in the first health monitoring log;
determining one or more selected phrases for the first health monitoring log including at least one of a first number of relevant phrases from the one or more relevant phrases that are most frequently used in the first health monitoring log and a second number of relevant phrases from the one or more relevant phrases that are most distinctively used in the first health monitoring log relative to other journal entries in the one or more other journal entries; and
determining the one or more extracted symptoms based on the one or more selected phrases.

### Claim 4 (depends on 3)

The computer-implemented method of claim 3, wherein determining the one or more extracted symptoms based on the one or more selected phrases comprises:
for each selected phrase of the one or more selected phrases, determining a corresponding extracted symptom of the one or more extracted symptoms.

### Claim 5 (depends on 1)

The computer-implemented method of claim 1, wherein determining the second distribution of extracted sentiment scores over the first period of time comprises:
determining, for each health monitoring log of the one or more health monitoring logs, one or more sentiment indicators;
identifying, for each time unit of one or more time units associated with the first period of time, one or more first health monitoring logs of the one or more health monitoring logs associated with the time unit;
determining, for each time unit of the one or more time units, one or more first sentiment scores, wherein the one or more first sentiment scores for a time unit of the one or more time units includes each one or more sentiment indicators associated with a first health journal of the one or more first health journals that is associated with the time unit; and
combining each one or more first sentiment scores associated with a time unit of the one or more time units to determine the first distribution.

### Claim 6 (depends on 5)

The computer-implemented method of claim 5, wherein the one or more sentiment indicators for a first health monitoring log of the one or more health monitoring logs include one or more general sentiment indicators for the first health monitoring log and one or more behavioral health sentiment indicators for the first health monitoring log.

### Claim 7 (depends on 5)

The computer-implemented method of claim 5, wherein the one or more sentiment indicators for a first health monitoring log of the one or more health monitoring logs include one or more expressive sentiment indicators for the first health monitoring log and one or more internal health sentiment indicators for the first health monitoring log.

### Claim 8 (depends on 1)

The computer-implemented method of claim 1, wherein generating the aggregate distribution of inferred health-related predictions over the first period of time comprises generating a graph object that depicts, for each consecutive occurrence of a first symptom associated with the one or more health monitoring logs over one or more time units in the first period of time, a corresponding semantic score from the second distribution.

### Claim 9 (depends on 1)

The computer-implemented method of claim 1, wherein generating the second distribution comprises:
determining, for each occurrence of an extracted symptom on a first time unit of the one or more time units, a sentiment score based on one or more sentiment indicators.

### Claim 10 (depends on 1)

The computer-implemented method of claim 1, wherein generating the second distribution comprises:
determining, for all occurrences of an extracted symptom on a first time unit of the one or more time units, a sentiment score based on one or more sentiment indicators.

### Claim 11 (depends on 1)

The computer-implemented method of claim 1, further comprising generating one or more medical diagnoses based on the aggregate distribution.

### Claim 12 (depends on 1)

The computer-implemented method of claim 1, further comprising generating one or more medical recommendations based on the aggregate distribution.

### Claim 13

An apparatus comprising at least one processor and at least one memory including program code, the at least one memory and the program code configured to, with the processor, cause the apparatus to at least:
determine, based on the one or more health monitoring logs, a first distribution of symptomatic prediction labels over a first period of time associated with the one or more health monitoring logs;
process the one or more health monitoring logs and using a sentiment detection machine learning model to determine a second distribution of extracted sentiment scores over the first period of time;
generate, based on the first distribution and the second distribution, an aggregate distribution of inferred health-related predictions over the first period of time; and
provide a presentation for display via a user interface, wherein the presentation comprises the aggregate distribution.

### Claim 14 (depends on 13)

The apparatus of claim 13, wherein determining the first distribution comprises:
determining, for each health monitoring log of the one or more health monitoring logs, one or more extracted symptoms;
identifying, for each time unit of one or more time units associated with the first period of time, one or more first health monitoring logs of the one or more health monitoring logs associated with the time unit;
determining, for each time unit of the one or more time units, one or more first extracted symptoms, wherein the one or more first extracted symptoms for a time unit of the one or more time units include each one or more extracted symptoms for a first health monitoring log of the one or more first health monitoring logs that is associated with the time unit; and
combining each one or more first extracted symptoms associated with a time unit of the one or more time units to determine the first distribution.

### Claim 15 (depends on 14)

The apparatus of claim 14, wherein determining the one or more extracted symptoms for a first health monitoring log of the one or more journal entries comprises:
identifying one or more relevant phrases in the first health monitoring log;
determining one or more selected phrases for the first health monitoring log including at least one of a first number of relevant phrases from the one or more relevant phrases that are most frequently used in the first health monitoring log and a second number of relevant phrases from the one or more relevant phrases that are most distinctively used in the first health monitoring log relative to other journal entries in the one or more other journal entries; and
determining the one or more extracted symptoms based on the one or more selected phrases.

### Claim 16 (depends on 15)

The apparatus of claim 15, wherein determining the one or more extracted symptoms based on the one or more selected phrases comprises:
for each selected phrase of the one or more selected phrases, determining a corresponding extracted symptom of the one or more extracted symptoms.

### Claim 17

A computer program product comprising at least one non-transitory computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions configured to:
determine, based on the one or more health monitoring logs, a first distribution of symptomatic prediction labels over a first period of time associated with the one or more health monitoring logs;
process the one or more health monitoring logs and using a sentiment detection machine learning model to determine a second distribution of extracted sentiment scores over the first period of time;
generate, based on the first distribution and the second distribution, an aggregate distribution of inferred health-related predictions over the first period of time; and
provide a presentation for display via a user interface, wherein the presentation comprises the aggregate distribution.

### Claim 18 (depends on 17)

The computer program product of claim 17, wherein determining the first distribution comprises:
determining, for each health monitoring log of the one or more health monitoring logs, one or more extracted symptoms;
identifying, for each time unit of one or more time units associated with the first period of time, one or more first health monitoring logs of the one or more health monitoring logs associated with the time unit;
determining, for each time unit of the one or more time units, one or more first extracted symptoms, wherein the one or more first extracted symptoms for a time unit of the one or more time units include each one or more extracted symptoms for a first health monitoring log of the one or more first health monitoring logs that is associated with the time unit; and
combining each one or more first extracted symptoms associated with a time unit of the one or more time units to determine the first distribution.

### Claim 19 (depends on 18)

The computer program product of claim 18, wherein determining the one or more extracted symptoms for a first health monitoring log of the one or more journal entries comprises:
identifying one or more relevant phrases in the first health monitoring log;
determining one or more selected phrases for the first health monitoring log including at least one of a first number of relevant phrases from the one or more relevant phrases that are most frequently used in the first health monitoring log and a second number of relevant phrases from the one or more relevant phrases that are most distinctively used in the first health monitoring log relative to other journal entries in the one or more other journal entries; and
determining the one or more extracted symptoms based on the one or more selected phrases.

### Claim 20 (depends on 19)

The computer program product of claim 19, wherein determining the one or more extracted symptoms based on the one or more selected phrases comprises:
for each selected phrase of the one or more selected phrases, determining a corresponding extracted symptom of the one or more extracted symptoms.
