Humanoid Robot Dialogue System Personalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current dialogue methods for humanoid robots are repetitive, stereotyped, and unnatural, failing to establish a genuine emotional bond with human interlocutors, as they separate information acquisition and usage phases and lack adaptability to individual users.
Innovation Solution
A method that identifies human interlocutors, extracts and updates their profiles with dialogue variables, analyzes spontaneous sentences, and formulates responses using multiple sentence models and rules, allowing for personalized and progressive dialogue that integrates information acquisition and usage seamlessly, enabling natural and non-stereotyped interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional dialogue methods are used, then the dialogue structure is simple and easy to implement, but the dialogue becomes repetitive and stereotyped
Solution Approach 1:
The dialogue system is segmented into multiple independent sentence models (MDP1, MDP2, etc.), each representing different dialogue patterns. These models are organized in a hierarchical structure with syntactic trees, allowing the system to select appropriate models based on the interlocutor profile without requiring a completely complex system architecture.
Solution Approach 2:
The interlocutor profile is built and updated in advance during the dialogue, storing dialogue variables that characterize the user. This preliminary action allows the system to adapt subsequent responses based on accumulated knowledge, resolving the contradiction between simple structure and adaptive behavior.
2Ease of operation
If information acquisition and usage are separated into distinct phases, then the dialogue structure is clear and manageable, but the dialogue becomes unnatural
Solution Approach 1:
The system merges information acquisition and usage into a unified process. Sentence models contain both acquisition patterns (for extracting new information) and usage patterns (for formulating responses) within the same structural framework. This allows the dialogue to flow naturally while maintaining manageable complexity through the standardized model structure.
3Adaptability or versatility
If a single sentence model is used, then the system is simple and fast, but it cannot handle diverse dialogue situations
Solution Approach 1:
The system dynamically selects which sentence model to apply based on the current dialogue context and interlocutor profile. Rather than using a fixed single model or exhaustively checking all models, the system adapts its behavior in real-time by choosing the most appropriate model for the situation, balancing versatility with operational simplicity.
4Adaptability or versatility
If dialogue variables are not stored in an interlocutor profile, then the system uses less memory, but the dialogue cannot be personalized
Solution Approach 1:
The system stores only the specific dialogue variables that are relevant to personalization in the interlocutor profile, rather than storing all possible information. Each sentence model is designed to extract and store only the locally relevant variables needed for that particular dialogue pattern, optimizing memory usage while maintaining personalization capability.
Data Source
Figure 1
Figure 2
Figure 3A~3D
AI summary
The invention relates to a method for dialogue between a machine (OE), preferably a humanoid robot (R), and at least one human interlocutor (INT). Said method comprises the following steps, implemented by said computer: a) identifying said human interlocutor; b) extracting, from a database (BDD), an interlocutor profile (INT_PR) including a plurality of dialogue variables, at least one value being attributed to at least one of said dialogue variables; c) receiving and analyzing at least one phrase from said interlocutor; and d) formulating and sending at least one response phrase on the basis of at least said phrase, received and interpreted during Step c), and a so-called "dialogue variable" from said interlocutor profile.