User-Specific Data Integration for Large Language Model Answers

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Large-scale language models struggle to provide answers based on user-specific information not generally available, as they are trained on publicly accessible data, limiting their ability to respond accurately to user-specific queries.

Innovation Solution

An information processing system that integrates a large-scale language model with user-specific databases, including conversation history and in-company information, to generate answers tailored to individual users by using prompts that leverage this information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If a large-scale language model is used to generate answers, then the model can process a large amount of data and generate comprehensive responses, but the model cannot access user-specific information that is not generally available

Engineering Contradiction:

Improveamount of dataVSAvoiduser-specific information

Core Design Contradiction:

Quantity of substanceVSLoss of information

Solution Approach 1:

The patent introduces an intermediary mechanism that retrieves user-specific information from external sources (databases, APIs) and incorporates it into the prompt before sending to the large-scale language model. This intermediary layer enables the model to access user-specific data without changing the model itself, resolving the contradiction between using a general-purpose model and accessing private user information.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system segments the information processing into separate components: general knowledge from the pre-trained model and user-specific information from external sources. By segmenting the data sources and processing steps, the patent allows the large-scale language model to focus on general pattern recognition while user-specific details are handled separately through retrieval and integration mechanisms.

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If user-specific information is integrated into prompts, then answers become more relevant and accurate, but the prompt length increases which may exceed token limits

Engineering Contradiction:

Improveanswer accuracyVSAvoidprompt length

Core Design Contradiction:

Measurement precisionVSLength of moving object

Solution Approach 1:

The system applies partial action by selectively retrieving and incorporating only the necessary user-specific information relevant to the current query, rather than including all available user data. This selective approach maintains answer accuracy while controlling prompt length within token limits.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The patent applies local quality by tailoring the prompt to include user-specific information only where relevant to the specific question being asked. Instead of uniformly adding user data to all prompts, the system identifies and incorporates user-specific details locally at points where they directly impact the answer relevance and accuracy.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS20250272319A1Information processing apparatus, information processing system, information processing method, and recording medium

Publication Date: 2025.08.28 RICOH CO LTD

AI summary

An information processing apparatus includes circuitry that receives an input of question data, outputs to a large-scale language model a prompt requesting generation of answer data to the question data based on user-specific information, and outputs to a terminal apparatus output answer data that is output from the large-scale language model in response to an input of the prompt.