Ensemble Chatbot Architecture for Diverse Conversational Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conversational systems are siloed, knowledge-inhibited, and require slow manual training, limiting their ability to handle diverse conversations and increasing time to market for new classes of interactions.
Innovation Solution
A universal conversational system employing an ensemble of chatbots, including NER, knowledge graph-based, and generative models, with a supercharged transformer architecture and quality assurance function to generate responses, leveraging the wisdom of the crowd for continuous learning and improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual training is used for conversational systems, then training precision can be controlled, but training time increases significantly
Solution Approach 1:
The system segments the training process into multiple specialized chatbot components (NER chatbot, knowledge graph-based chatbot, generative chatbot) that can be trained independently and parallelized, reducing overall training time while maintaining precision through targeted training for each component
Solution Approach 2:
The system performs preliminary actions by pre-training multiple specialized chatbot models before deployment. These pre-trained models can be quickly assembled and fine-tuned for specific tasks, eliminating the need for extensive manual training from scratch and significantly reducing time to market
2Reliability
If rule-based or knowledge model approaches are used, then conversational quality can be maintained, but the system becomes siloed and cannot handle diverse conversations
Solution Approach 1:
The system merges multiple specialized chatbot models (NER, knowledge graph-based, and generative chatbots) into a unified ensemble architecture. This combination allows the system to leverage the strengths of each approach while handling diverse conversation types, achieving both quality and versatility through collaborative processing
Solution Approach 2:
The ensemble chatbot system achieves universality by integrating multiple functional components that can handle different types of conversations. The system can adapt to various domains and conversation styles through its multi-functional architecture, breaking down the silos of traditional rule-based systems while maintaining quality through coordinated processing
3Adaptability or versatility
If ensemble of multiple chatbots is used, then conversational quality and versatility improve, but system complexity increases
Solution Approach 1:
The system introduces a quality assurance function as an intermediary that manages the ensemble of chatbots. This mediator component coordinates the multiple chatbot models, aggregates their outputs, and selects the best responses, simplifying the overall system architecture while maintaining the benefits of the ensemble approach
Data Source
AI summary
Conversational systems are intelligent machines that can understand language and conversing with a customer in writing or verbally. Embodiments herein provide a method for generating a universal conversational system using an ensemble of chatbots and a universal conversational system that adopts wisdom of crowd manifesting as an ensemble of chatbots. The ensemble of chatbots takes responses from NER and rule based conversational models. The knowledge based conversation models where complex queries that require question and answer, and the ensemble of generative knowledge chatbots are relying on a pre-trained models. The pre-trained models are complemented by domain specific training to answer queries that fall outside rule-based chatbot or knowledge graph-based conversation bot capability. The universal conversational system capable of building online virtuous automated learning loop where the models learn from each other and also from human response as wisdom of crowd.