Character-Style Response Generation With Context Retrieval

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing chatbot systems face challenges in mimicking the conversational styles of fictional characters due to difficulties in defining characters with few descriptions, representing discrete styles, and obtaining sufficient dialogue data for training.

Innovation Solution

Utilizing a response generation method that acquires utterance data, selects a context from a candidate set using a retrieval model, and generates dialogue sets with a language model to mimic the character's style, leveraging large-scale language models through pseudo-dialogue prompting and bi-encoders.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional chatbot methods are used to generate responses, then the system can output predetermined responses, but it cannot effectively mimic the conversational styles of fictional characters

Engineering Contradiction:

Improveability to mimic fictional character stylesVSAvoidaccuracy of style representation

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system performs preliminary actions by acquiring utterance data and selecting relevant contexts from a candidate set before generating responses. This preparation enables the language model to accurately mimic fictional character styles by having pre-processed and organized training data that captures the desired conversational patterns.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system copies the conversational style of fictional characters by using utterance data and selected contexts as templates for generating responses. The language model learns to replicate the style patterns from the provided utterance data, allowing the chatbot to respond like a specific fictional character rather than using generic responses.

Inventive Principle:

Principle #26Copying

2Ease of manufacture

If discrete style labels are used to define characters, then the system can categorize character types, but it cannot capture the nuance and complexity of actual character conversations

Engineering Contradiction:

Improveease of character definitionVSAvoidprecision of style representation

Core Design Contradiction:

Ease of manufactureVSManufacturing precision

Solution Approach 1:

The system extracts the essential style information from fictional character conversations by selecting relevant contexts from a candidate set based on utterance data. This extraction process captures the nuanced conversational patterns without requiring discrete style labels, allowing the system to maintain both ease of definition and high precision of representation.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system changes the approach from using discrete style labels to using continuous utterance data and context selection. By transforming the representation from categorical labels to data-driven context matching, the system achieves both ease of character definition and precise style representation through the language model's ability to process and generalize from the extracted contexts.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If large amounts of dialogue data are collected for training, then the model can learn accurate character styles, but the data collection process becomes time-consuming and complex

Engineering Contradiction:

Improveaccuracy of style mimicryVSAvoidtime required for data collection

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system extracts only the necessary style information from fictional character conversations by selecting contexts from a candidate set that are most relevant to the utterance data. This extraction approach maintains high accuracy in style mimicry while significantly reducing the amount of data that needs to be collected and processed, thereby saving time in the data collection process.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The system uses partial action by selecting only the most relevant contexts from the candidate set rather than processing all available data. This selective approach achieves sufficient accuracy for style mimicry without requiring exhaustive data collection, balancing the trade-off between data quantity and time investment.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20250384872A1Method of generating response using utterance and apparatus therefor

Publication Date: 2025.12.18 HYPERCONNECT INC

AI summary

Systems and techniques to generate imitative responses are illustrated. response generation method performed in an electronic apparatus of the present disclosure includes acquiring at least one piece of utterance data, acquiring a first context corresponding to the utterance data from a context candidate set, generating one or more dialogue sets including the first context and the utterance data, receiving a second context from a user, and acquiring a response corresponding to the second context using a language model based on the one or more dialogue sets.