SCH: Natural Language Processing for Enhanced Behavioral Counseling

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

As more people seek out counseling help, Natural Language Processing (NLP) technology can provide support for the growing number of counselor professionals to deliver quality-focused services. The overarching goal of this project is to make advances toward a new generation of NLP systems. It is expected to have significant implications in the way counseling is conducted because it will provide new ways to evaluate counselor effectiveness and assisting them with ongoing feedback and coaching in the form of automatic coding, as well as turn-taking and language suggestions. This will allow counselor professionals and other health care practitioners to improve the quality of their counseling through timely and cost-effective feedback. The methodology developed in this project will establish the foundations toward systems that can provide support for counseling interactions for a wide range of health care providers from physicians and nurses to disease management coaches and dietitians. The project will pursue several new and unique research directions in NLP inspired by the growing area of behavioral counseling. Specifically, the project targets the following four main objectives. (1) Create a large dataset of behavior counseling with extensive annotations, addressing several behaviors and covering several online and offline sources. (2) Develop methods that combine the recent advances in neural networks with symbolic representations encoding counseling strategies, and use these methods to classify counselor behaviors and predict their most likely future behavior based on previous interactions with the client. (3) Develop natural language generation models that will assist the counselors in their conversations, specifically focusing on the generation of questions and reflections. The methods will consist of generative neural models that learn from a large counseling dataset while explicitly modeling counseling strategies and integrating expert knowledge bases. (4) Create and evaluate a framework for the integration of NLP tools to provide feedback and coaching to counselors in training. Importantly, the project will incorporate feedback from domain experts and end users across the entire research pipeline, using focus groups to identify needs, preferences, and features to include in each design and implementation stage of the proposed research objectives. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.