Turn abandoned patents
into your own filing
into your own filing
This invention describes a method for analyzing text documents by first building a multi-level graph that connects the document to its paragraphs, sentences, and individual words. It then uses an attention model to assign importance weights to connections within this graph and applies a Graph Neural Network (GNN) to process the graph, updating the features of each text unit. Finally, it generates a document summary or other output for an NLP task, which can include highlighting key words, sentences, or paragraphs, or visualizing the graph with important nodes indicated.
Why it matters: The dramatic advancements in large language models and transformer architectures since 2020 provide significantly more powerful and accessible methods for generating contextual token features and understanding hierarchical text relationships, potentially enhancing the performance or simplifying the implementation of this graph-based NLP system.
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
