Sign in Subscribe
Machine learning

Hybrid Natural Language Processing System for Outputting Customized Responses to Natural Language Formatted Queries

Strategic Insights

— 1 min read

Researchers have filed a patent application for a hybrid natural language processing system that utilizes both a large language model and a dynamic, machine learning response application to generate customized responses to natural language formatted queries. The invention aims to address the limitations of conventional large language models, which often struggle to provide nuanced, specific, and real-time responses.

Key Takeaways:

  • The proposed system combines a large language model with a dynamic, machine learning response application to process natural language formatted queries.
  • The large language model receives the query and separates it into two portions: a stock portion and a set of real-time conditions.
  • The model then generates a response template that includes placeholders for real-time data, which are filled in by the dynamic response application.
  • The dynamic response application retrieves real-time data from private data stores, such as bank account data stores, and inputs this data into the response template.
  • The final response is customized for the specific user and time, and may be displayed on a hardware display, such as a computing device or smartphone.
  • The system aims to overcome the limitations of conventional large language models by utilizing real-time data and providing nuanced, specific, and real-time responses.
  • The patent claims outline a method for using a hybrid natural language processing system to output customized responses to natural language formatted queries.

Statistics:

  • The large language model uses one or more neural networks with a plurality of parameters (over one billion) trained on one or more large data stores.
  • The dynamic response application processes the natural language formatted query and the response template in real-time.
  • The system can identify one or more user accounts included in the set of real-time conditions and retrieve user account information from one or more private data stores.
  • The private data stores are continually updated in real-time, allowing the system to provide up-to-date responses.
  • The patent application supplements the background information with an inventors' summary information.

Sources:

  • Jhaveri, Rajan; Noorizadeh, Emad; Welles, Chris. Dynamic Response Engine. U.S. Patent Application Number 20250181615, filed February 3, 2025 and posted June 5, 2025.
  • Patent URL: https://ppubs.uspto.gov/pubwebapp/external.html?q=(20250181615)&db=US-PGPUB&type=ids

Read more