Conversational System Graph Update via NLP Intermediary
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conversational systems, such as chat robots, are limited by predefined conversation graphs and cannot provide desired answers when user questions fall outside the scope of the predefined questions and answers, leading to errors.
Innovation Solution
A method that updates the conversation graph with a second message sequence generated based on a knowledge library including expert knowledge associated with the conversation topic, allowing the system to handle situations not covered by the original graph.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a predefined conversation graph is used to structure dialogues, then the system can provide structured and manageable conversation flows, but it cannot provide desired answers when user questions fall outside the predefined scope
Solution Approach 1:
An NLP engine is introduced as an intermediary component between the user input and the conversation graph. The NLP engine processes user questions that do not match predefined conversation paths, extracts intent and entities, and generates appropriate responses or routes to relevant conversation nodes, thereby extending the system's ability to handle out-of-scope questions while maintaining the structured benefits of the conversation graph
Solution Approach 2:
The system transitions from a static conversation graph to a dynamic hybrid architecture where the conversation graph coexists with an NLP-based natural language processing component. This dynamic structure allows the system to adaptively switch between predefined conversation paths for known topics and generative NLP responses for unknown topics, enhancing versatility while preserving reliability for covered topics
2Adaptability or versatility
If the conversation graph is expanded to cover more topics, then the system can handle a wider range of questions, but the complexity of maintaining and updating the graph increases
Solution Approach 1:
Instead of directly expanding and maintaining a comprehensive conversation graph for all possible topics, the system uses the NLP engine to generate responses by copying and adapting from training data and knowledge bases. This allows the system to handle diverse topics without proportionally increasing the complexity of the conversation graph structure
Solution Approach 2:
The NLP engine serves as a universal component that can handle multiple topics and question types without requiring separate predefined conversation paths for each. This multi-functional approach allows the system to cover a wide range of topics while maintaining a relatively simple conversation graph structure for core dialogues
3Reliability
If a knowledge library with expert knowledge is integrated, then the system can generate accurate responses for unknown topics, but the system complexity and processing time increase
Solution Approach 1:
Expert knowledge and training data are pre-processed and stored in structured knowledge bases and language models during system initialization. The NLP engine leverages these pre-computed resources to generate responses quickly during runtime, avoiding the need for real-time complex reasoning while maintaining high response accuracy
Solution Approach 2:
The system applies NLP-based generative responses selectively only when conversation graph matching fails or confidence is low, rather than processing all user inputs through the full NLP pipeline. This partial application approach maintains response accuracy for uncertain cases while minimizing processing time for routine conversations that match the conversation graph
Data Source
AI summary
Embodiments of the present invention relate to computer-implemented methods, systems, and computer program products for managing a conversational system. In one embodiment, a computer-implemented method comprises: obtaining, by a device operatively coupled to one or more processors, a first message sequence comprising messages involved in a conversation between a user and a conversation server; obtaining, by the device, a conversation graph indicating an association relationship between messages involved in a conversation; and in response to determining that the first message sequence is not matched in the conversation graph, updating, by the device, the conversation graph with a second message sequence, the second message sequence being generated based on a knowledge library including expert knowledge that is associated with a topic of the conversation.