LLM Response Generation Using Vectorized Conversation History
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems lack the ability to effectively utilize previous interaction content and contextual attributes to generate tailored and accurate responses during real-time conversations with customers, leading to inefficiencies in customer service interactions.
Innovation Solution
A host platform utilizing large language models (LLMs) to identify contextual attributes from conversations, convert them into vectorized data, and store them in a database for retrieval and analysis, enabling generation of customized responses based on previous interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional communication systems are used without LLMs and contextual analysis, then the system complexity remains low, but the response accuracy and personalization are insufficient
Solution Approach 1:
The system segments the complex task of generating accurate responses into multiple components: contextual attribute identification, vectorized data generation, database storage, and response generation. Each component is handled by a specialized module (LLM executor, vectorized data generator, database manager), allowing the system to achieve high response accuracy while managing complexity through modular architecture
Solution Approach 2:
The patent introduces intermediary components including the vectorized data generation module that transforms contextual attributes into structured vector representations, and the database manager that intermediates between stored interaction data and the response generation process. These intermediaries bridge the gap between raw data and accurate responses, enabling precision without direct complex processing at each stage
2Adaptability or versatility
If previous interaction content is stored and analyzed using LLMs, then the response personalization improves, but the processing time and computational resources increase
Solution Approach 1:
The system performs preliminary actions by pre-processing interaction content into vectorized data representations and storing them in the database before they are needed for response generation. The vectorized data generation module prepares contextual attributes in advance, and the database manager organizes this data for quick retrieval, eliminating the need for time-consuming processing during actual communication sessions
Solution Approach 2:
The patent creates simplified copies of complex interaction data by generating vectorized representations of contextual attributes. Instead of storing and processing raw unstructured interaction content, the system creates compact vector copies that capture essential meaning and can be quickly retrieved and processed by the LLM executor, maintaining personalization while reducing processing time
3Loss of information
If contextual attributes are identified and stored in a database, then the information retrieval capability improves, but the data storage requirements and system complexity increase
Solution Approach 1:
The system changes the parameter of data representation by transforming contextual attributes into vectorized form. This parameter change compresses rich contextual information into efficient vector representations that can be stored compactly in the database while maintaining full retrieval capability, reducing storage requirements without losing information
Data Source
AI summary
An example operation may include one or more of receiving interaction content from a communication session between a source device and a service provider device of a service provider, identifying a search criteria from the interaction content, retrieving a subset of vectors from a plurality of vectors stored in a vector database based on the search criteria of the interaction content, wherein the subset of vectors includes previous interaction content with the service provider, generating a response for the communication session based on execution of a large language model (LLM) on the subset of vectors, and outputting the response to at least one of the source device and the service provider device during the communication session. The example operation may further include an AI agent that performs an action based on the response.