This invention describes a method for building the knowledge base for a voice-activated question-answering system. It focuses on creating a network of questions that are semantically similar, meaning they have the same underlying meaning even if phrased differently. This network is populated by defining initial questions and their answers, then expanding them using semantically related words like synonyms, often sourced from linguistic databases like WORDNET.
Why it matters: Filed before the advent of modern large language models, the manual or rule-based generation of semantically variant questions and populating a speech lattice would have been a significant challenge. Today, advanced AI and LLMs can automate and vastly improve the semantic understanding and generation required for this system's core functionality.
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