This invention describes an automated companion that can hold a conversation with a user. It works by taking in various types of information from the user, like their voice, what they see, text, and even touch, to understand their current state and the conversation's context. Based on this understanding, along with a pre-set conversation structure and lessons learned from past dialogues, the system then figures out and delivers an appropriate response. The claims specify that the input includes audio, visual, text, and haptic data, which are analyzed for content, emotion, and surrounding information.
Why it matters: Filed before large language models could generate nuanced dialogue. Modern multimodal AI and LLMs can now extract features and determine responses with greater sophistication.
AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.
Reinvent this with AI