# US 2019/0172449 A1 · Methods and systems for optimized selection of data features for a neuro-linguistic cognitive artifical intelligence system

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

- Number: US 2019/0172449 A1
- Document ID: US20190172449A1
- Status: Published application
- Kind: A1
- Application number: 16/120,943
- Filed: 2018-09-04
- Published: 2019-06-06
- Earliest claimed: 2016-04-06
- Assignees: Omai Ai, Inc.
- Applicants: Omni Ai, Inc.
- Inventors: GANG XU; Tao YANG; Ming-Jung SEOW
- Web page: https://patentgrep.com/patent/US20190172449A1

## CPC

- G10L 15/16: using artificial neural networks
- G06F 40/237: Lexical tools
- G06N 3/042: Knowledge-based neural networks; Logical representations of neural networks
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/088: Non-supervised learning, e.g. competitive learning
- G06N 20/00: Machine learning
- G06V 20/52: Surveillance or monitoring of activities, e.g. for recognising suspicious objects (recognising microscopic objects G06V20/69)
- G06V 40/20: Movements or behaviour, e.g. gesture recognition (recognition of facial expressions G06V40/16)
- G10L 15/197: Probabilistic grammars, e.g. word n-grams
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Techniques are disclosed to optimize feature selection in generating betas for a feature dictionary of a neuro-linguistic Cognitive AI System. A machine learning engine receives a sample vector of input data to be analyzed by the neuro-linguistic Cognitive AI System. The neuro-linguistic Cognitive AI System is configured to generate multiple betas for each of a plurality of sensors. The machine learning engine identifies a sensor specified in the sample vector and selects optimization parameters for generating betas based on the identified sensor.

## Claims

### Claim 1

A computer-implemented method to optimize feature selection to generate an adaptive linguistic model for a cognitive Artificial Intelligence (AI) system, the method comprising:
receiving from a sensor, a sample vector of input data, the sample vector of input data indicating a type of the sensor;
identifying via at least one processor, the sensor based on the sample vector of input data;
generating via the at least one processor, a plurality of feature symbols by organizing the sample vector of input data into probabilistic clusters;
determining via the at least one processor, optimization parameters based on the sensor identified;
generating via the at least one processor, a plurality of betas based at least in part on at least one of combinations of the plurality of feature symbols or combinations of the optimization parameters; and
generating via the at least one processor, an adaptive linguistic model based at least in part on combinations of the plurality of betas, the adaptive linguistic model configured for a cognitive AI system to distinguish between normal and abnormal activity in received input data.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein the sample vector of input data is a first sample vector of input data, the method further comprising:
receiving, from the sensor, a second sample vector of input data;
generating, via the at least one processor, a plurality of betas for the second sample vector of input data; and
updating via the at least one processor, the adaptive linguistic model based at least in part on the plurality of betas generated for the second sample vector of input data.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein determining the optimization parameters further comprises:
tuning a plurality of parameters based on the type of the sensor, the adaptive linguistic model being generated at least in part on the plurality of parameters.

### Claim 4 (depends on 3)

The computer-implemented method of claim 3, wherein the tuning is based on at least one of: a maximum length of each feature word in the plurality of betas, a maturity threshold for each feature word in the plurality of betas, a statistical significance of each feature word in the plurality of betas, and/or a feature-combination rule set; and the feature-combination rule set is generated from the plurality of betas.

### Claim 5 (depends on 1)

The computer-implemented method of claim 1, wherein the sensor is at least one of an image sensor, a video sensor, an audio sensor, or a SCADA sensor.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein the optimization parameters differ based on the type of the sensor.

### Claim 7 (depends on 1)

The computer-implemented method of claim 1, wherein generating the adaptive linguistic model includes determining a tunable strategy to generate the plurality of betas based on the type of the sensor.

### Claim 8

A non-transitory computer-readable storage medium storing instructions, that when executed by a computer system, perform operations to generate an adaptive linguistic model, the instructions comprising instructions to:
identify a sensor based on a sample vector of input data, the sample vector of input data obtained from the sensor and indicating a type of the at least one sensor;
generate a plurality of feature symbols by organizing the sample vector of input data into probabilistic clusters;
determine, after the identification, optimization parameters based on the sensor;
generate a plurality of betas based at least in part on at least one of combinations of the plurality of feature symbols or combinations of the optimization parameters; and
generate the adaptive linguistic model based at least in part on combinations of the plurality of betas, the adaptive linguistic model configured for use in a cognitive artificial intelligence system to distinguish between normal and abnormal activity in received input data.

### Claim 9 (depends on 8)

The computer-readable storage medium of claim 8, wherein the sample vector of input data is a first sample vector of input data, the instructions further comprising instructions to:
identify the sensor based on a second sample vector of input data;
generate, after identification, a plurality of betas for the second sample vector of input data; and
update the adaptive linguistic model based at least in part on the plurality of betas generated for the second sample vector of input data.

### Claim 10 (depends on 8)

The computer-readable storage medium of claim 8, wherein determining the optimization parameters further comprises instructions to:
tune a plurality of parameters based on the type of the sensor, the adaptive linguistic model being generated at least in part on the plurality of parameters.

### Claim 11 (depends on 10)

The computer-readable storage medium of claim 10, wherein the tuning is based on at least one of a maximum length of each feature word in the plurality of betas, a maturity threshold for each feature word in the plurality of betas, a statistical significance of each feature word in the plurality of betas, and/or a feature-combination rule set; and wherein the feature-combination rule set is generated from the plurality of betas.

### Claim 12 (depends on 8)

The computer-readable storage medium of claim 8, wherein the sensor is at least one of an image sensor, a video sensor, an audio sensor, or a SCADA sensor.

### Claim 13 (depends on 8)

The computer-readable storage medium of claim 8, wherein the instructions to generate the adaptive linguistic model includes instructions to determine a tunable strategy to generate the plurality of betas based on the type of the sensor.

### Claim 14 (depends on 8)

The computer-readable storage medium of claim 8, wherein the optimization parameters differ based on the type of the sensor.

### Claim 15

A system, comprising:
a processor;
and a memory, including an application program configured to perform operation for processing data, the operations comprising:
identifying a sensor based on a sample vector of input data, the sample vector of input data obtained from the sensor and indicating a type of the at least one sensor;
generating a plurality of feature symbols by organizing the sample vector of input data into probabilistic clusters;
determining, after the identifying, optimization parameters based on the sensor;
generating a plurality of betas based at least in part on at least one of combinations of the plurality of feature symbols or combinations of the optimization parameters;
generating the adaptive linguistic model based at least in part on combinations of the plurality of betas, the adaptive linguistic model configured for distinguishing between normal and abnormal activity in received input data.

### Claim 16 (depends on 15)

The system of claim 15, wherein the sample vector of input data is a first sample vector of input data, the operations further comprising:
identifying the sensor based on a second sample vector of input data;
generating, after identifying, a plurality of betas for the second sample vector of input data; and
updating the adaptive linguistic model based at least in part on the plurality of betas generated for the second sample vector of input data.

### Claim 17 (depends on 15)

The system of claim 15, wherein determining the optimization parameters further comprises:
tuning a plurality of parameters based on the type of the sensor, the adaptive linguistic model being generated at least in part on the plurality of parameters.

### Claim 18 (depends on 17)

The system of claim 17, wherein:
the tuning is based on at least one of a maximum length of each feature word in the plurality of betas, a maturity threshold for each feature word in the plurality of betas, a statistical significance of each feature word in the plurality of betas, or a feature-combination rule set,
the feature-combination rule set is generated from the plurality of betas.

### Claim 19 (depends on 15)

The system of claim 15, wherein generating the adaptive linguistic model includes determining a tunable strategy to generate the plurality of betas based on the type of the sensor.

### Claim 20 (depends on 15)

The system of claim 15, wherein the optimization parameters differ based on the type of the sensor.