mental health banner


aiberry’s innovative technology is the result of over a decade of research performed by our Chief Scientist Dr. Newton Howard, who is one of today’s foremost experts in computational and cognitive neurosciences and his collaborators. This in-depth research in the field of Sentiment Analysis, Intention Awareness and Multi-modal Dialogue Systems positions us to create novel and impactful behavioral health products and platforms. 

Approach Towards a Natural Language Analysis for Diagnosing Mood Disorders and Comorbid Conditions

Dr. Newton Howard (2013)

An approach for developing a diagnosis system for mood disorders, such as depression and bipolar disorder, based on language analysis from speech and text. Our system is based on the Mood State Indicator algorithm (MSI) for real-time analysis of a patient’s mental state. MSI is designed to give a quantitative measure of cognitive state based on axiological values and time orientation of lexical features. MSI’s multi-layered analytic engine consists of multiple information processing modules to systematically retrieve, parse and process features of a patient’s discourse.

Emotion Recognition in Conversation: Research Challenges, Datasets, and Recent Advances

Soujanya Poria, Navonil Majumder, Rada Mihalcea, Eduard Hovy (2019)

Emotion is intrinsic to humans and consequently emotion understanding is a key part of human-like artificial intelligence (AI). Emotion recognition in conversation (ERC) is becoming increasingly popular as a new research frontier in natural language processing (NLP) and is extremely important for generating emotion-aware dialogues that require an understanding of the user’s emotions. ERC has tremendous potential in health-care systems as a tool for psychological analysis.

Fusing audio, visual and textual clues for sentiment analysis from multimodal content

Soujanya Poria, Erik Cambria, Newton Howard, Guang-Bin Huang, Amir Hussain (2015)

A novel methodology for multimodal sentiment analysis, which consists in harvesting sentiments from videos by leveraging audio, visual and textual modalities as sources of information. The approach uses both feature-and decision-level fusion methods to merge affective information extracted from multiple modalities. Preliminary comparative experiments show that the proposed multimodal system achieves an accuracy of nearly 80%, outperforming all state-of-the-art systems.

Real-time vocal features extraction for automated suicidal risk assessment in hotlines

C. Aguet, L. Falissard, F. Shurmann, N. Howard

Mental health disorders, and more particularly depression and suicide, have become a major concern for society. Conventional diagnosis suffers from a lack of accuracy and reliability due to patient self-reporting. Public health interventions include suicide crisis hotline . However, counsellors answering the calls are not necessarily mental health professionals and can overlook subtle signs. The human voice carries much more than simply the linguistic content. Paralinguistic content, including information regarding emotions, is also present. As a person’s mental state indirectly alters his or her speech. The purpose of this research is to evaluate suicidal risk in hotline calls through speech analysis and give real-time feedback regarding the caller’s suicidal impulse

Towards a Differential Diagnostic of PTSD Using Cognitive Computing Methods

Newton Howard, Louis Jehel, Romain Arnal (2014)

The aim of this research is to construct a predictive algorithm based on linguistic computation to predict Post Traumatic Stress Disorder (PTSD) and comorbid conditions. Most contemporary work in psychometrics, concerns itself with the construction and validation of instruments in an attempt to measure personality, attitudes, sentiment and beliefs. Measurement of these phenomena is difficult and largely subjective. Our approach analyses natural language using the Mind State Indicator algorithm (MSI). MSI may be able to diagnose PTSD and specify differential diagnosis from comorbid disorders. Future work using this method could potentially predict PTSD symptom severity and treatment-related changes in these symptoms.

Get your download by filling out the form below: