Turn abandoned patents
into your own filing
into your own filing
This invention describes a way to understand relationships between things mentioned in a piece of text. It works by breaking down the text into individual parts, like the specific items and the words describing their connection. Each of these parts, along with related information, gets turned into a unique numerical code called an embedding. These codes are then combined into a larger, composite code, which is fed into an artificial intelligence system to identify the relationships.
Why it matters: Filed before the widespread adoption of large language models (LLMs), this invention's core task of generating sophisticated text embeddings and extracting relationships has been revolutionized. Modern LLMs can now perform these functions with significantly greater accuracy and less custom engineering than was possible in 2019, making the implementation of such a system potentially more efficient and powerful today.
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
