Escalation detection using sentiment analysis

Patent metadata

Field	Value
Publication number	US-10224059-B2
Application number	US-201715812325-A
Country	US
Kind code	B2
Filing date	Nov 14, 2017
Priority date	Jul 21, 2016
Publication date	Mar 5, 2019
Grant date	Mar 5, 2019

Abstract

Systems an methods of computer-based escalation detection are disclosed. In embodiments, a method includes: determining an occurrence of an interaction event between a first and second party within a recording including audio data; analyzing the audio data; determining an escalation during the interaction event to generate escalation data; partitioning each interaction event into a plurality of sections; assigning a sentiment score for each of the plurality of section; calculating an overall sentiment score for the interaction event, wherein the saved escalation data includes the overall sentiment score; graphing the overall sentiment score for the interaction event with a plurality of additional overall sentiment scores from respective additional interaction events to produce a graph of sentiment scores over time; and detecting a pattern of increasing of decreasing overall sentiment scores over time based on the graph.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
determining, by a computer device, an occurrence of an interaction event between a first party and a second party within a recording including audio data;
analyzing, by the computer device, the audio data;
determining, by the computer device and based on the analyzing the audio data, an escalation during the interaction event to generate escalation data;
saving, by the computer device, the escalation data
partitioning, by the computer device, each interaction event into a plurality of sections, wherein a first section represents a start of the interaction event, and another section represents an end of the interaction event;
assigning, by the computer device, a sentiment score for each of the plurality of sections;
calculating, by the computer device, an overall sentiment score for the interaction event by combining the sentiment scores for each of the plurality of sections, wherein the saved escalation data includes the overall sentiment score;
graphing, by the computer device, the overall sentiment score for the interaction event with a plurality of additional overall sentiment scores from respective additional interaction events to produce a graph of sentiment scores over time; and
detecting, by the computer device, a pattern of increasing or decreasing overall sentiment scores over time based on the graph.

2. The computer-implemented method of claim 1 , wherein determining the occurrence of an interaction event includes detecting, by the computing device, a threshold level of sound in the audio data.

3. The computer-implemented method of claim 1 , further comprising comparing, by the computing device, the escalation data to historic escalation data.

4. The computer-implemented method of claim 1 , further comprising:
gathering, by the computing device, overall sentiment scores for a plurality of interaction events involving the first party; and
monitoring, by the computing device, the overall sentiment scores over time to determine any increase or decrease in the overall sentiment scores.

5. The computer-implemented method of claim 1 , further comprising:
converting, by the computing device, the audio data into text,
wherein the analyzing the audio data comprises analyzing the text for a change in tone over time.

6. The computer-implemented method of claim 5 , wherein the audio data as a whole is converted to text and analyzed without distinguishing between audio data generated by the first party and audio data generated by the second party.

7. A computer program product for detecting escalation the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computing device to cause the computing device to:
determine an occurrence of an interaction event between a first party and a second party within an audio-video recording including audio data;
analyze the audio data for a change in tone over time;
analyze the audio data for a presence of any negative tones;
determine whether the change in tone, the presence of any negative tones, or a combination of the change in tone and the presence of any negative tones, indicates an escalation during the interaction event to generate escalation data;
save the escalation data;
compare the saved escalation data to historic escalation data;
gather overall sentiment scores for a plurality of interaction events involving the first party;
monitor the overall sentiment score over time to determine any increase or decrease in the overall sentiment scores; and
graph the overall sentiment scores, wherein the monitoring comprises detecting a pattern of increasing overall sentiment scores over time based on the graph.

8. The computer program product of claim 7 , wherein the program instructions to determine the occurrence of an interaction event include instructions to detect a threshold level of sound in the audio data.

9. The computer program product of claim 7 , wherein the recording is a video recording and the program instructions to determine the occurrence of an interaction event include instructions to detect that the first party or the second party has entered a frame of the video recording for a predetermined time period.

10. The computer program product of claim 7 , the program instructions further causing the computing device to:
partition each interaction event into a plurality of sections, wherein a first section represents a start of the interaction event, and another section represents an end of the interaction event;
assign a sentiment score for each of the plurality of sections based on the changes in the tone over time and the presence of any negative tones; and
calculate an overall sentiment score for the interaction event by combining the sentiment scores for each of the plurality of sections, wherein the saved escalation data includes the overall sentiment score.

11. The computer program product of claim 7 , the program instructions further causing the computing device to convert the audio data into text, wherein the instructions to analyze the audio data for a change in tone over time and the instructions to analyze the audio data for the presence of any negative tones comprises analyzing the text.

12. The computer program product of claim 11 , wherein the audio data as a whole is converted to text and analyzed without distinguishing between audio data generated by the first party and audio data generated by the second party.

13. A system for detecting escalation, comprising:
a CPU, a computer readable memory and a computer readable storage medium associated with a computing device;
program instructions to determine an occurrence of a first interaction event between a first person of interest and one or more members of the public within an audio-video recording including audio data and video data;
program instructions to analyze the audio data for a change in tone over time;
program instructions to analyze the audio data for a presence of any negative tones;
program instructions to determine whether the change in tone, the presence of any negative tones, or a combination of the change in tone and the presence of any negative tones, indicates an escalation during the first interaction event to generate escalation data;
program instructions to save the escalation data;
program instructions to gather overall sentiment scores for a plurality of interaction events involving the first party;
program instructions to monitor the overall sentiment scores over time to determine any increase or decrease in the overall sentiment scores;
program instructions to partition each interaction event into a plurality of sections, wherein a first section represents a start of the first interaction event, and another section represents an end of the first interaction event;
program instructions to assign a sentiment score for each of the plurality of sections based on the changes in the tone over time and the presence of any negative tones;
program instructions to calculate an overall sentiment score for the first interaction event by combining the sentiment scores for each of the plurality of sections, wherein the saved escalation data includes the overall sentiment score;
program instructions to compare the overall sentiment score with an overall sentiment score of a second interaction event involving a second person of interest; and
program instructions to rate the first and second persons of interest based on the comparing the overall sentiment scores,
wherein the program instructions are stored on the compu

Assignees

• IBM

Inventors

• Childress Rhonda L
• Eckert Kim A
• Mcnair Ryan D

Classifications

• G10L25/63 Primary

  for estimating an emotional state · CPC title

• G10L25/57

  for processing of video signals · CPC title

• G06Q10/105

  Human resources · CPC title

• G10L15/26

  Speech to text systems (G10L15/08 takes precedence) · CPC title

• G06Q30/016

  After-sales · CPC title