Speech Analysis

Emotion detection using speech analysis by Nemesysco

Speech Analysis - Emotion detection using speech analysis by Nemesysco

Amir Liberman, CEO of Nemesysco, to Speak at Haifa Technion on the May 26, 2014.

The age of the emotional machines – A  look into emotion detection technologies, its practical uses today and a glance into the world of tomorrow

Abstract
The Layered Voice Analysis emotion detection technology is based on unique methods of signal processing and deals with challenges from the technical point of view, but just as importantly it raises questions on the psychological and theoretical aspects of human emotions. What makes our emotions unique? Are the basic emotions really basic? Can machines really “understand” emotions, and maybe more importantly –  Do we really want them to?

Determining the Veracity of a Claim | PropertyCasualty360

Wouldn’t you like to know whether a claim is true and accurate on the front end of the claims process? Claims fraud is an epidemic these days and determining the veracity of a claim early in the process is imperative. Veremetrix has a claim fraud detection solution, developed by Nemesysco Ltd, in Israel, that analyzes incoming or outgoing claims calls and quickly and accurately detects fraud up front in real time. The solution is called RA7 and….

via Determining the Veracity of a Claim | PropertyCasualty360.

Isolated populations emotional monitoring

Using Nemesysco’s voice analysis for Emotional state monitoring of isolated populations

People in isolated facilities (prisons, distant working locations, army bases, ships etc…) are exposed to many different types of events and forces that have a severe impact on their emotional state, and may drive them to extreme actions, harming others or themselves. In open facilities, the effect of such emotional charge may be dissolved by detachment from the group, a friendly companion or a pleasurable act. However, the isolated facility offers very limited possibilities for extreme emotions to be dissolved, and in fact increasing the need for immediate identification and affirmative action by local supervisors.

The emotional changing forces include external influences such as other inmates or peers acts and disregards, conversations with remote parties and management level attention. Additionally, internal forces develop over time out of memories and new sets of personal expectations not necessarily related to external effects.

Monitoring internal verbal communication among the isolated population, as well as monitoring their communication with the external world using voice analysis may provide important indications about the person’s state of mind, and allow proper attention and treatment. Additionally, the system can provide some predictive value to near-future actions – and assist the prevention of self-harm, suicide and assault on others.

Few challenges are easily identified:

  1. The collection of the relevant voice data may not be trivial, simply because some people just don’t communicate much with others
  2. Some events may be planned and carried out in a short time frame and with no audio data containing relevant signals (no recorded session is available following the “action decision”)
  3. Action by proxy – Some events may be planned by one but carried out by others – and so pinpointing the individual making a decision may not be enough

Despite those limitations, voice monitoring system can provide useful data for decision makers and treatment managers and will even perform much better if connected to additional sources of information. Integration with other evaluation systems, databases containing past emotional records (thus helping to observe trend development and critical changes) and inner-group connections map (that may help to identify different roles and relationships) will enhance the system’s capabilities dramatically.

Using Nemesysco’s security level voice analysis technology, voice data from any source and /or communication system (such as internal and external phone system) is automatically processed and associated with the relevant person. Databases can be updated with new emotional data and predictive algorithms are then used to identify critical change and issue a warning immediately. Trend analysis is used to identify a problematic trend development over time, and unique emotional cues can be associated with different relationships between the individuals and the outside world.

Using certain emotional parameters it will be easier to identify communication containing key-words, action commands and other illegal activities, and assist the work of the intelligence officers in the relevant facilities.

The unique abilities of Nemesysco’s LVA technology can assess many different emotional components such as stress, anger, positive and negative emotional reactions, concentration, anticipation and much more, thereby creating very profound datasets for advanced analysis.

* Nemesysco is currently in first field tests of SCA1 based emotional state analysis system for inmates in the USA

 

Directorate of Forensic Sciences to offer agricultural, rural forensics – Times Of India

“Eariler, DFS introduced technologies such as teleforensics, suspect identification system, layered voice analysis, forensic narco-analysis tools, Integrated Ballistics Identification System (IBIS), 3D scanner for crime scenes and disciplines such as wildlife forensics. Another significant collaboration during the summit will be formation of a centre of excellence at GFSU by Israel-based Nemesysco Ltd that provided the DFS one of the country’s first Layered Voice Analysis (LVA) technologies.

The centre will be used by researchers and professionals from across the globe for research and development.

via Directorate of Forensic Sciences to offer agricultural, rural forensics – Times Of India.

The New LVA-i movie clip

Dear all,

Check out Nemesysco’s new video clip for LVA-i pre-employment and vetting system here.

LVA-i balances between high availability, optional accuracy and a fair price tag per test. See how easy it is to use LVA-i and how flexible the system really is.

To learn more about LVA-i, please visit LVA-i official website at: www.lva-i.com

 

 

 

Info Request QA5








CRM News: Customer Service: Voice of the Customer: Analyzing What People Say and How They Say It

How people speak is as important as the words they say when it comes to sentiment analysis. “The human voice is made of many repeating patterns,” said Amir Liberman, CEO of Nemesysco. “Everything that attracts your attention will generate a slight disturbance in these patterns. We look carefully at this voice chart, and we identify disturbances.”

via CRM News: Customer Service: Voice of the Customer: Analyzing What People Say and How They Say It.

Microscope of Emotions- The speakers “Emotional Diamond” research

Emotional Diamond

LVA Technology: The Emotional Diamond

30/11/2012

Netanya

Effective Speakers under the Emotional Microscope: Commons and Differences of Lecturers’ Emotional Patterns of their Speech and the “Emotional Diamond”

 

Internal study, seeking academic research partners!

Introduction

The fact that lecturers and politicians alike have commons and differences in their way of speech was the key idea for this research – but what makes one speech (or speaker) more effective than another?

Following a request from TEDxAmsterdam to analyze the emotional profile of the speakers of 2012, we took TED[1] as a great platform for a research on this topic. Our main research assumption is that the way of speech and its emotional content can be analyzed using Nemesysco’s LVA and its Emotional Detection platform (EDP), and as such it can give a new level of knowledge pertaining to how speakers transform their knowledge, beliefs, and experience in an  unconsciousness way.

We choose Nemesysco’s QA5 call center technology as the basis for this research, due to its automatic nature and the variety of emotional patterns (not veracity) which can be evolved during a lecture or a political speech. In this research we deliberately exclude the veracity assessment element as it is not in the scope of research.

With the use of the QA5 analysis tool, we can measure different levels of energy, stress, excitement, concentration, passion, embarrassment and other parameters among our tested speakers.

We’ve also made an assumption, based on some preliminary tests we made, that yet another critical factor for successful presentation (beyond the dedication of the speaker to the content) is the matter of training to reach the ability of actually enjoying the course of a lecture, promoting the message without hesitation, in full confidence and energy.

Research Analyzing Tool – The Emotional Diamond™

We have customized the LVA technology based product, QA5[2], (often used for monitoring emotional states of customer speech during an on-going calls in call centers) and added the “Emotional Diamond” display to allow easy reading of many layers of emotional data. A new report display was added as well to assist and quantify the presence of the various key emotions (see below), but for the purpose of complete research protocol, all QA5 regular parameters will be used.

Clinton's Emotional Diamond

President Clinton’s Emotional Diamond analysis – Passionate and energetic

Obama's Emotional Diamond

Obama’s speech – “Super stars” speakers exhibit high energy and low thinking

Method

In order to analyze the emotional patterns, we have chosen several emotions that usually stand for contrasts comparison. We configured and defined a table of attributes that can be calculated and translated into visual and textual conclusions.
The following table of Attributes clarifies the concept that we should use various parameters to demolish the paradigm on one hand, and to build a solid base of clear evidence to the assumption, on the other hand. This table comprises of the 3 groups of parameters: Technical, Rate of emotional states and Emotional states. Each group concludes relevant parameters to be focused on, since we can assume that those have an effect upon the results and the conclusions to be taken.

Technical Definition
Time The duration of the speech/analyzed portion
CPA QA5’s “Call Priority” – a function summarizing negative emotions in the analyzed voice assuming the voice belongs to a customer
FCP The FINAL Call Priority value at the end of the voice segment
AP QA5’s Agent Priority score – a function summarizing the over all negative emotions assuming the tested voice belongs to the service provider.
Rate of Emotional states
%anger Percentage of voice segments in which ANGER was detected
%High Stress Percentage of voice segments in which high levels of STRESS were detected
%Upset Percentage of voice segments in which UPSET feelings was detected
%Content Percentage of voice segments in which happiness or joy were detected
Emotional/ Cognitive States
Av. Energy Average level of “Emotional energy” as detected in the call
Energy Indicates how much energy evolves in the tested party effort to deliver
Passion Indicates how passionate the tested party is, with regard to the content he delivers
Emotional Indicates the emotional state of the tested party
Shy Indicates how shy (embarrassed) the tested party is
Stress Indicates how nervous the tested party is.
Thinking Indicates the mental effort of the tested party
Cog Stress Indicates the level of certainty/cognitive stress  the tested party experience
Concentration Indicates how concentrated the tested party is
Anticipation Indicates how anticipated the tested party is for reaction

Speech Analysis Results for the Speakers’ profiles

As part of our initial research 8 recorded speeches from different speakers were analyzed, each for its entire duration.
Therefore, since we initiated this research in small scale it is important to point out the conclusion should be taken as initial results only. It should further be assumed that to properly classify a true “Speaker’s Profile”, several speeches for each speaker should be analyzed.
Nevertheless, the results of each voice test presented us with interesting measurements that used to configure 3 profiles of speakers.


The “Politician” / “idealistic” Profile – The “Superstar lecturers”

In theory, such speaker can be defined as the one who delivers a message he believes in, not always relevant to current situation, the one who has a vision.
According to the test results, the complex of measures is:
High percentage of content                          (16.87-17.39)
High rate of average energy                        (12.84-17.64)
High rate of energy                                         (67-78)
High rate of passion state                             (74)
High rate of emotion state                            (64-72)
Low rate of shy state                                      (37)
Low rate of thinking state                           (19)
Low rate of cognitive stress                         (14-16)
High rate concentration state                     (31)
Medium rate of anticipation state              (28-40)

Profile structure: Concentrated in the goal that should be reached, passionate beliefs, knows what to say and doesn’t have to think a lot, high drive to reach a summit, low stress. (based on the analysis of Bill Clinton and Barack Obama)


The “professional” Profile
In theory, such as a speaker can be defined as the one who has a speech plan. Know exactly what to say, well-practiced in such a mode of speaking, saves energy, knows how to deliver the state of a speech (not always what defines the same speaker in a normal way of speaking), plays a role that serves the cause, controls the emotional state to the minimum in order to gain points and results derived from his speech.

High percentage of anger                             (0.91)
High percentage of upset                           (4.95)
Low rate of average energy                         (7.41)
Low rate of energy                                        (43)
Low rate of passion state                             (37)
Low rate of stress state                                (34)
Low rate of concentration state                 (26)
High rate of anticipation state                    (53)

Profile structure: Cool. Delivers a role in a play, saves energy for the entire speech (aka monologue), creates and delivers anticipation.  (Based on the analysis of Benjamin Netanyahu UN speech and Winston Churchill’s “Never before” speech)

The “true story teller”
In theory, such as a speaker brings a story with a role within. That could be a scientific case study, narrative story, family story or other kind. Sometimes, there can be little inaccuracy in involved facts, but there is only a need to bring it to story to the audience.

High percentage of stress                             (26.67-82.09)
High percentage of upset                           (27.17-32.84)
High percentage of content                       (16.87-17.39)
Low rate of average energy                         (3.81-4.24)
Low rate of energy                                        (22-26)
Low rate of passion state                             (20)
Low rate of emotion state                           (44-51)
High rate of shy state                                    (55-56)
High rate of stress state                                (46-64)
High rate of thinking state                            (32-34)
High rate of cognitive stress                        (29-36)
High rate of concentration state                (32-33)
Medium rate of anticipation state              (38-41)

Profile structure: Such speaker knows exactly what the subject is. Speaker as such, lives the story for real. Anxious and Excited. (The tested parties are 3 TED lecturers)

 

Conclusions

This table of results concludes the above parameters. Note that the results, are based on one speech for each speaker, and in order to reach a solid conclusion, it is advisable to first, make a comparison of several speeches for each speaker on his own. Mainly, since the emotional state can be changed sometimes, as a result of environmental terms, type of audience, reason to deliver the speech, etc.

References

  1. http://www.ted.com/pages/about
  2. http://www.qa5system.com/
  3. John G. Wacker, A definition of theory: research guidelines for different theory-building research methods in operations management, Journal of Operations Management 16 1998 361–385
  4. Susan Dugdale, How to write a speech – step by step help: http://www.write-out-loud.com/howtowritespeech.html
  5.  Roddy CowieRandolph R. Corneliu, Describing the emotional states that are expressed in speech, Speech Communication, Volume 40, Issues 1–2, April 2003, Pages 5–32

 

 

 

Contact us: bizdev@nemesysco.com

Website: www.nemesysco.com

 



[1]  TED is a nonprofit devoted to Ideas worth spreading. It started out in 1984 as a conference.

[2] QA5 offers a standardized set of measurements to properly evaluate the agents’ performance and customers’ satisfaction, automatically, and with minimal hardware resources consumption. The aggregated data is used for automated report generation and standardizing your KPI (Key performance indicators).