Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Toward a Reinforcement-Learning-Based System for Adjusting Medication to Minimize Speech Disfluency

Pavlos Constas, Vikram Rawal, Matthew Honorio Oliveira, Andreas Constas, Aditya Khan, Kaison Cheung, Najma Sultani, Carrie Chen, Micol Altomare, Michael Akzam, Jiacheng Chen, Vhea He, Lauren Altomare, Heraa Murqi, Asad Khan, Nimit Amikumar Bhanshali, Youssef Rachad, Michael Guerzhoy

TL;DR

The paper tackles the problem of reducing speech disfluency linked to mental health by proposing a hypothetical reinforcement learning system that automatically adjusts medications. It combines a disfluency-detection pipeline (Whisper ASR → DT-ACNN tagging → GPT-2 fine-tuning) with a patient-simulation and LinUCB-based RL framework to optimize medication regimens. Results from a 500-run simulation indicate that the RL approach can achieve meaningful fluency improvements in a plausible setting, with a 52% success rate and average fluency of 0.66/1.00, though substantial variability and possible failures highlight the need for careful modeling and safety considerations. The work demonstrates the feasibility of automatic data collection and data-driven medication optimization for speech disfluency, while acknowledging regulatory, ethical, and clinical hurdles that must be overcome before real-world deployment.

Abstract

We propose a reinforcement learning (RL)-based system that would automatically prescribe a hypothetical patient medication that may help the patient with their mental health-related speech disfluency, and adjust the medication and the dosages in response to zero-cost frequent measurement of the fluency of the patient. We demonstrate the components of the system: a module that detects and evaluates speech disfluency on a large dataset we built, and an RL algorithm that automatically finds good combinations of medications. To support the two modules, we collect data on the effect of psychiatric medications for speech disfluency from the literature, and build a plausible patient simulation system. We demonstrate that the RL system is, under some circumstances, able to converge to a good medication regime. We collect and label a dataset of people with possible speech disfluency and demonstrate our methods using that dataset. Our work is a proof of concept: we show that there is promise in the idea of using automatic data collection to address speech disfluency.

Toward a Reinforcement-Learning-Based System for Adjusting Medication to Minimize Speech Disfluency

TL;DR

The paper tackles the problem of reducing speech disfluency linked to mental health by proposing a hypothetical reinforcement learning system that automatically adjusts medications. It combines a disfluency-detection pipeline (Whisper ASR → DT-ACNN tagging → GPT-2 fine-tuning) with a patient-simulation and LinUCB-based RL framework to optimize medication regimens. Results from a 500-run simulation indicate that the RL approach can achieve meaningful fluency improvements in a plausible setting, with a 52% success rate and average fluency of 0.66/1.00, though substantial variability and possible failures highlight the need for careful modeling and safety considerations. The work demonstrates the feasibility of automatic data collection and data-driven medication optimization for speech disfluency, while acknowledging regulatory, ethical, and clinical hurdles that must be overcome before real-world deployment.

Abstract

We propose a reinforcement learning (RL)-based system that would automatically prescribe a hypothetical patient medication that may help the patient with their mental health-related speech disfluency, and adjust the medication and the dosages in response to zero-cost frequent measurement of the fluency of the patient. We demonstrate the components of the system: a module that detects and evaluates speech disfluency on a large dataset we built, and an RL algorithm that automatically finds good combinations of medications. To support the two modules, we collect data on the effect of psychiatric medications for speech disfluency from the literature, and build a plausible patient simulation system. We demonstrate that the RL system is, under some circumstances, able to converge to a good medication regime. We collect and label a dataset of people with possible speech disfluency and demonstrate our methods using that dataset. Our work is a proof of concept: we show that there is promise in the idea of using automatic data collection to address speech disfluency.
Paper Structure (25 sections, 3 equations, 4 figures, 2 tables)

This paper contains 25 sections, 3 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Data Collection
  3. Methods
  4. Rating System
  5. Rater Performance Analysis
  6. Data Model
  7. Analysis
  8. Standardized Ratings
  9. Disfluency Pipeline
  10. Transcribing Audio with Whisper
  11. Disfluency Tagging
  12. Fine-tuning GPT-2
  13. Results
  14. Patient Simulation and Reinforcement Learning
  15. Overview
  16. ...and 10 more sections

Figures (4)

  • Figure 1: Learning curves: disfluency prediction regression task
  • Figure 2: A diagram of a Hidden Markov Model representing context emissions generated from states.
  • Figure 3: Convergence to higher speech fluency.
  • Figure 4: Lack of convergence to higher speech fluency.