Behavioral AI Labs logo

Behavioral AI Labs works with Fortune500 to unleash the power hidden in their audio and text data, to enhance their customers’ and employees’ communication and decision making. We work towards your eureka! moment and then we shorten the path from value discovery to service deployment. Our agile US-based team of Ph.D.s is trained on executing continuously on the discover-plan-build-deploy cycle, significantly shortening time to market, using our wealth of unique data resources, machine learning expertise, and Behavioral Signal’s AI platform.

Behavioral Signals Whitepaper Fig.1
Behavioral AI Labs subplot

Many R&D outfits out there claim to work on machine learning, natural language processing, speech recognition, voice-of-the-customer, etc. What makes us unique is our ability to detect signals from audio data associated with behaviors and traits that drive human decision making. Our decisions are conditioned by who we are (personality, morality), how we feel (emotion, mood, regulation), how we are perceived socially (status, dominance, raport), our goals and self-awareness (persuasion, awareness, attention). As the behavioral signal detection capabilities reach maturity and our emotion AI technology approaches human performance unique business opportunities arise. Understanding not only what you say but how you say it and what you mean is our secret sauce.

Some of the Projects We’ve Worked On

Customer Engagement


Movies and TV

Voice Assistants/Robotics

Voice training software for sales reps

Behavioral profiling for your customers and sales reps

Real-time goal-oriented feedback for your sales force

Behavioral Therapy: Training and quality assurance for psychotherapists and patients

Wellness: Track mood swings using audio signals captured on device

Couples Therapy: Predicting the status of relationship change from vocal cues

Computational media intelligence for diversity and inclusion in movies.
“It’s only a matter of time where the formula of what makes up a great story, a great character can be learned by a computer” —MacDonald, LA Times, April 11, 2019

Voice training software for dubbing

Engagement and Gamification for voice assistants via tracking voice tone

Real-time emotion detection for voice assistants

4 Steps to Value Deployment





During a 2h consultation, with senior researchers, we will discuss the problem you are trying to solve and create an assessment of the business opportunity

The next stage includes a 2 to 3-week technology feasibility study and initial experimentation providing an estimate of effort, data, and time needed to achieve the goal. Additionally, we will be able to review initial performance indicators and associated RoI. We will also define the API needed by the application developers.

The 3rd step involves development, in over a 2 to 3-month process, including the following phases:
i) data collection/annotation (if needed), feature and model engineering, performance optimization and validation,
ii) connect to the data store,
iii) build and test APIs

Produced APIs or MVPs are put to production and end-to-end performance is evaluated and monitored

Our solutions are based on open-source software and on the leading emotion and behavioral AI platform from Behavioral Signals and owned by the customer. Your data privacy is assured using only HIPPA/GDPR compliant cloud resources under your control.

R&D Leads

Prof. Shrikanth Narayanan, USC – Scientific Lead

Theodoros Giannakopoulos, II&T, NCSR Demokritos – ML Lead

Nassos Katsamanis, ATHENA – Technical Lead

Select Publications & Awards

Deep Hierarchical Fusion with application in Sentiment Analysis —Georgiou, Papaioannou, Potamianos, 2019

6 time winner of the INTERSPEECH quality of human interactions & computational paralinguistics challenge

Data Augmentation using GANs for Speech Emotion Recognition —Chatziagapi, Paraskevopoulos, Sgouropoulos, Pantazopoulos, Nikandrou, Giannakopoulos, Katsamanis, Potamianos, Narayanan, 2019

Winner: Gold-standard Emotion Sub-challenge at the 2018 ACM Audio-Visual Emotion Challenge

Behavioral Signal Processing: Deriving Human Behavioral Informatics From Speech
and Language —Narayanan, Georgiou, 2013

Winner: Sentiment analysis twitter challenge SemEval/NAACL 2016

Our motto (to paraphrase Ekman): Emotions and thoughts determine the quality of our life while behaviors and actions determine the outcomes in our life. Behavioral Signal Processing technology can transform your business!

Contact us today for an initial conversation labs @