Developing AI Conversation Technology for Comfortable, Human-Centric Interaction

Sharp has developed an AI conversation technology that realizes comfortable interactions with human-centric responses through an automated evaluation system for AI response content. Analysis of conversation trends when using Sharp's AI-equipped products and services revealed that interactions go beyond product functions. Believing that enjoying casual conversations leads to attachment to products and services, the company focused on developing AI conversation technology as part of its proprietary CE-LLM AI technology. Previously, evaluating AI conversation responses relied on subjective assessment, leading to issues such as time consumption and inconsistency among evaluators. Furthermore, because there were no unified standards or indicators to comprehensively evaluate the 'desirability' of conversations, quantitative verification and evaluation were difficult. To address these issues, the company conducted research on prior studies regarding conversations, extracted and systematized items that affect the 'desirability' of conversations, such as 'responsiveness,' 'context understanding,' and 'knowledge,' and constructed evaluation criteria that can quantitatively measure response content.