# US 2020/0401768 A1 · Detecting anomolies in textual items using cross-entropies

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

- Number: US 2020/0401768 A1
- Document ID: US20200401768A1
- Status: Published application
- Kind: A1
- Application number: 16/903,916
- Filed: 2020-06-17
- Published: 2020-12-24
- Earliest claimed: 2019-06-18
- Assignees: Verint Americas Inc.
- Applicants: Verint Americas Inc.
- Inventors: Cynthia Freeman
- Web page: https://patentgrep.com/patent/US20200401768A1

## CPC

- G06F 40/30: Semantic analysis
- G06F 40/47: Machine-assisted translation, e.g. using translation memory
- G06F 40/216: using statistical methods
- G06F 40/279: Recognition of textual entities
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

In an implementation, a method for detecting anomalies in textual items is provided. The method includes: receiving a first plurality of textual items by a computing device; training a language model using the received first plurality of textual items by the computing device; after training the language model, receiving a second plurality of textual items by the computing device; calculating a cross-entropy for each textual item in the second plurality of textual items by the computing device using the language model; and detecting an anomaly in at least one of the textual items of the second plurality of textual items by the computing device using the calculated cross-entropies.

## Claims

### Claim 1

A method for detecting anomalies in textual items comprising:
receiving a first plurality of textual items by a computing device;
training a language model using the received first plurality of textual items by the computing device;
after training the language model, receiving a second plurality of textual items by the computing device;
calculating a cross-entropy for each textual item in the second plurality of textual items by the computing device using the language model; and
detecting an anomaly in at least one of the textual items of the second plurality of textual items by the computing device using the calculated cross-entropies.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
determining that a number of textual items in the second plurality of textual items exceeds a threshold; and
in response to the determination, sliding the language model forward using the second plurality of textual items.

### Claim 3 (depends on 1)

The method of claim 1, further comprising:
determining that an amount of time has passed since the language was last slid forward exceeds a threshold; and
in response to the determination, sliding the language model forward using the second plurality of textual items.

### Claim 4 (depends on 1)

The method of claim 1, wherein a textual item comprises a plurality of words.

### Claim 5 (depends on 4)

The method of claim 4, wherein calculating the cross-entropy for each textual item in the second plurality of textual items comprises calculating the cross-entropy using a subset of the plurality of words of the textual item.

### Claim 6 (depends on 1)

The method of claim 1, wherein calculating the cross-entropy for each textual item in the second plurality of textual items comprises calculating an average cross-entropy for each textual item in the second plurality of textual items.

### Claim 7 (depends on 1)

The method of claim 1, wherein the textual item comprises a call transcript between an agent and a customer, a chat transcript between the agent and the customer, or messages in a social networking application.

### Claim 8 (depends on 1)

The method of claim 1, further comprising:
receiving a list of textual items that were determined not to be associated with anomalies;
for each textual item of the second plurality of textual items:
determining if the textual item is in the list of textual items; and
if the textual item is in the list of textual items, reducing the calculated cross-entropy for the textual item.

### Claim 9 (depends on 8)

The method of claim 8, wherein determining if the textual item is in the list of textual items comprises determining if the textual item is similar to a textual item in the list of textual items.

### Claim 10

A method for detecting anomalies in textual items comprising:
receiving a first plurality of textual items by a computing device;
training a language model using the received first plurality of textual items by the computing device;
after training the language model, receiving a second plurality of textual items by the computing device;
for each textual item of the second plurality of content items:
calculating a cross-entropy for the textual item using the language model by the computing device;
calculating a perplexity of the language model by the computing device; and
if the calculated perplexity exceeds a threshold, sliding the language model forward using a subset of the textual items of the second plurality by the computing device; and
detecting an anomaly in at least one textual item of the second plurality of textual items by the computing device using the calculated cross-entropies.

### Claim 11 (depends on 10)

The method of claim 10, wherein a textual item comprises a plurality of words.

### Claim 12 (depends on 10)

The method of claim 10, wherein calculating the cross-entropy for the textual item comprises calculating the cross-entropy using a subset of the plurality of words of the textual item.

### Claim 13 (depends on 10)

The method of claim 10, wherein calculating the cross-entropy for the textual item comprises calculating an average cross-entropy.

### Claim 14 (depends on 10)

The method of claim 10, wherein the textual item comprises a call transcript between an agent and a customer, a chat transcript between the agent and the customer, or messages in a social networking application.

### Claim 15 (depends on 10)

The method of claim 10, further comprising:
receiving a list of textual items that were determined not to be associated with anomalies;
for each textual item of the second plurality of textual items:
determining if the textual item is in the list of textual items; and
if the textual item is in the list of textual items, reducing the calculated cross-entropy for the textual item.

### Claim 16 (depends on 15)

The method of claim 15, wherein determining if the textual item is in the list of textual items comprises determining if the textual item is similar to a textual item in the list of textual items.

### Claim 17

A system for detecting anomalies in textual items comprising:
at least one processor; and
a memory storing instructions that when executed by the at least one processor cause the at least one processor to:
receive a first plurality of textual items;
train a language model using the received first plurality of textual items;
after training the language model, receive a second plurality of textual items;
calculate a cross-entropy for each textual item in the second plurality of textual items using the language model; and
detect an anomaly in at least one of the textual items of the second plurality of textual items using the calculated cross-entropies.

### Claim 18 (depends on 17)

The system of claim 17, wherein the instructions further comprise instructions that when executed by the at least one processor cause the at least one processor to:
determine that a number of textual items in the second plurality of textual items exceeds a threshold; and
in response to the determination, slide the language model forward using the second plurality of textual items.

### Claim 19 (depends on 17)

The system of claim 17, wherein the instructions further comprise instructions that when executed by the at least one processor cause the at least one processor to:
determine that an amount of time has passed since the language was last slid forward exceeds a threshold; and
in response to the determination, slide the language model forward using the second plurality of textual items.

### Claim 20 (depends on 17)

The system of claim 17, wherein a textual item comprises a plurality of words.