Conversational AI With Multi-Level Intent Routing for Deviations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional conversational AI systems struggle to imitate human conversations effectively, particularly in handling deviations and rephrasing, and managing transitions between sub-conversation units, leading to complex configurations and inconsistencies.
Innovation Solution
A system and method that classify user utterances into intents and configure conversations as sub-conversation units, allowing transitions based on previous outcomes, with multiple intent levels and localized strategies to generate responses, and connect to databases for additional information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional conversational AI systems use single-level intent classification, then the system structure remains simple, but the system fails to handle deviations and rephrasing effectively
Solution Approach 1:
The patent divides the conversational AI system into multiple sub-conversation units, each handling specific intents or aspects of the conversation. This segmentation allows the system to process deviations and rephrasing in dedicated units without overwhelming the entire system, thereby improving reliability while managing complexity through modular architecture.
Solution Approach 2:
The patent introduces a hierarchical dimension to intent classification by implementing multiple levels (first-level, second-level, and third-level classifiers). This dimensional transformation from a single-level to a multi-level classification system enables the AI to handle complex deviations and rephrasing by progressively narrowing down intents through multiple classification stages.
2Reliability
If the system uses multiple intent classification levels, then the ability to handle conversation deviations improves, but the configuration complexity increases
Solution Approach 1:
By segmenting the classification task into multiple levels with specific responsibilities (first-level for broad categories, second-level for sub-intents, third-level for specific actions), the patent makes the configuration more manageable. Each level can be configured and tested independently, reducing the overall configuration complexity despite the multi-level structure.
Solution Approach 2:
The patent introduces intermediate classification layers that act as mediators between user input and final response generation. These intermediate levels simplify the configuration by breaking down complex classification logic into smaller, more manageable steps, where each level handles a specific aspect of intent recognition.
3Adaptability or versatility
If the system allows free transitions between sub-conversation units, then the conversational flexibility improves, but the system loses ability to maintain desired path
Solution Approach 1:
The patent implements dynamic transition mechanisms between sub-conversation units that adapt based on conversation context, user intent, and system state. This dynamic approach allows the system to maintain flexibility for necessary deviations while automatically returning to the desired conversation path when appropriate, balancing adaptability with path maintenance.
Solution Approach 2:
The system incorporates feedback mechanisms that monitor conversation flow and detect when deviations from the desired path become excessive or unproductive. This feedback enables the system to gently guide the conversation back to the intended trajectory while still allowing legitimate user-initiated topic changes, maintaining both flexibility and direction.
4Measurement precision
If the system uses localized strategies for each sub-conversation unit, then the precision of information extraction improves, but the overall system complexity increases
Solution Approach 1:
The patent applies localized extraction strategies tailored to each sub-conversation unit's specific requirements and context. Each unit can employ extraction methods optimized for its particular intent type, improving precision for that specific context. The modular design ensures that this localized complexity is contained within individual units rather than affecting the entire system.
Data Source
AI summary
A system and a method to create a conversational artificial intelligence is disclosed. The presented system is uniquely designed to drive more human-like but yet robust conversations via text or voice. In at least one embodiment, the disclosed system is implemented on at least one computing device that can respond to at least one communicating entity which is hereinafter referred to as the user. The system can be configured to drive a conversation in any knowledge domain. The system disclosed herein, uses natural language processing techniques, natural language synthesis techniques and a novel strategy to respond to user inputs. The novel strategy may include a fundamental model of human conversation, an algorithm to correlate user inputs to the aforesaid model, an algorithm to handle questions from the user, an algorithm to avoid undesired inputs from the user and finally an algorithm to predict the next response generated by the system.