Scholarlink

To enhance communication aids driven by code-modulated visual evoked potentials (cVEP), selecting the right language model (LM) is paramount. With the new release of GPT4 turbo, this study compares “gpt-3.5-turbo-0125” (GPT-3.5T) and “gpt-4-turbo-2024-04-09” (GPT-4T). Given the increased time per token for GPT-4T compared to GPT-3.5T, the hypothesis posits that the operational characteristics of GPT-3.5T may better suit the unique demands of cVEP communication aids. By examining parameters such as user experience during BCI spelling in terms of Output Characters per Minute (OCM), accuracy and cost per session, this study aims to identify the most effective LM model for advancing cVEP-based communication aids.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparative Analysis of ChatGPT-3.5 Turbo and GPT-4 Turbo for Communication Aid in cVEP-Based BCI Speller Systems

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Atilla Cantürk,
  • Hannah Herschel,
  • Ivan Volosyak

摘要

To enhance communication aids driven by code-modulated visual evoked potentials (cVEP), selecting the right language model (LM) is paramount. With the new release of GPT4 turbo, this study compares “gpt-3.5-turbo-0125” (GPT-3.5T) and “gpt-4-turbo-2024-04-09” (GPT-4T). Given the increased time per token for GPT-4T compared to GPT-3.5T, the hypothesis posits that the operational characteristics of GPT-3.5T may better suit the unique demands of cVEP communication aids. By examining parameters such as user experience during BCI spelling in terms of Output Characters per Minute (OCM), accuracy and cost per session, this study aims to identify the most effective LM model for advancing cVEP-based communication aids.