LLM Document Conversation Interface with Relevance Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems lack an effective method to enable users to interactively converse with a corpus of electronic documents using large language models, failing to efficiently extract and present relevant information in a user-friendly interface.
Innovation Solution
A system and method that utilize a large language model to process user queries, determine relevant documents, and present them in a user interface, allowing users to select documents, enter queries, and receive relevant information, with the system comprising hardware processors, client computing platforms, and components for document retrieval, relevance determination, and interface presentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If users directly query a corpus of electronic documents using traditional search methods, then information can be retrieved, but the system cannot efficiently extract and present relevant information in a user-friendly interface
Solution Approach 1:
The patent introduces a large language model as an intermediary between the user's query and the document corpus. The LLM processes the query, identifies relevant documents, extracts key information, and presents it in a user-friendly format. This mediator resolves the contradiction by enabling both easy user interaction and efficient information extraction without requiring users to navigate complex search interfaces or manually analyze documents.
2Loss of information
If the system processes multiple documents through a large language model, then relevant information can be extracted, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the document processing task by first identifying and selecting only the most relevant documents based on the query, then processing those specific documents through the large language model. This segmentation approach reduces the number of documents that require complex LLM processing, thereby lowering computational complexity and processing time while still extracting all necessary relevant information.
3Loss of information
If the system provides comprehensive information from multiple documents, then information completeness is improved, but the risk of hallucinations increases
Solution Approach 1:
The patent implements a feedback mechanism where the system monitors and tracks the sources of information provided to users. By maintaining provenance information that links extracted content back to specific documents and passages, the system can verify the authenticity of information and reduce hallucinations. This feedback loop ensures information completeness while maintaining reliability through source verification.
4Ease of operation
If the system processes queries in real-time conversation mode, then user interaction is improved, but the processing speed and response time may be affected
Solution Approach 1:
The patent performs preliminary actions by pre-processing and indexing the document corpus before user queries are submitted. Documents are segmented, key features are extracted, and relevance criteria are established in advance. When users submit queries during conversation, the system can quickly match queries against pre-processed information, maintaining real-time interactive capabilities while ensuring fast response times.
Data Source
AI summary
Systems and methods for providing user interfaces to converse with a corpus of electronic documents via a large language model are disclosed. Exemplary implementations may: present a user interface configured to obtain entry of user input from a user to select one or more documents to be provided as input to a large language model for an individual conversation; responsive to selection of the individual conversation, provide an individual query as a prompt to the large language model; obtain and present an individual reply from the large language model; determine an individual document from the one or more documents that is relevant to the individual reply; present the individual document in a particular portion of the user interface; and/or perform other steps.