Dialogue System Using Causality Extraction for Topic Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dialogue systems face limitations in generating responses that develop topics from user input, as they are either confined to database scope, uncontrollable, or restricted to pre-defined scenarios, making it difficult to create engaging and expansive dialogues.
Innovation Solution
A dialogue apparatus and training device that utilize a neural network trained with causality expressions, allowing it to generate responses by extracting and chaining causality relationships from vast internet data, enabling the generation of related and latent results to user inputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a retrieval-based approach is used to generate responses from user input, then the response can be obtained from database, but the dialogue cannot develop beyond database scope and the relation between input and response is not clear to user
Solution Approach 1:
The patent extracts causality expressions from general internet data (news articles, documents) rather than relying on pre-existing dialogue databases. This allows the system to generate responses based on real-world causal relationships, enabling dialogue to develop beyond conventional database scope while maintaining reliability through factual causality-based responses.
2Adaptability or versatility
If deep learning-based end-to-end method is used to automatically generate responses, then the system can generate responses from user input, but the generated results are uncontrollable and the process is invisible
Solution Approach 1:
The patent introduces causality expressions as an intermediary representation between user input and generated response. Instead of direct black-box neural network generation, the system uses causality graphs as intermediaries that make the generation process transparent and controllable. Users can understand and control the response generation by examining the causality relationships, while still benefiting from flexible automatic generation.
3Ease of operation
If scenario-based approach is used for dialogue generation, then the response can be generated within prepared scenarios, but the dialogue is confined within limited boundaries
Solution Approach 1:
The patent creates a universal dialogue system that can handle multiple types of user inputs (questions, statements, commands) and generate appropriate responses across diverse topics using the same causality-based mechanism. The system extracts causality expressions from various internet sources and applies them universally to different dialogue scenarios, enabling both easy response generation and extensive topic coverage without requiring separate scenario preparations.
Data Source
AI summary
A training data generator and a training device include: a supposed input storage storing a plurality of supposed inputs supposed as inputs to a dialogue apparatus; expanded causality DB storing a plurality of causality expressions; a training data preparing unit extracting, for each of the plurality of supposed inputs stored in supposed input storage, a causality expression having a prescribed relation with said supposed input from the plurality of causality expressions, for forming a training data sample having the supposed input as an input and the extracted causality expression as an answer and storing in a training data storage; and a training unit training a response generating neural network designed to generate an output sentence to a natural language input sentence, by using the training data samples stored in training data storage.