Model-Based Conversational Agent for Complex Business Processes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current conversational virtual agents face limitations such as requiring large sets of tagged sentences for intent identification and entity extraction, struggling with complex business processes, inability to handle user interruptions, lack of support for business logic, and inability to maintain context and adapt dialog flows, as well as being modality-specific, making them unsuitable for natural human-like interactions.
Innovation Solution
A model-based intelligent conversational agent system that declaratively defines task models using a task modeling language, generates natural language grammar, and processes user inputs to provide context-aware and adaptable responses across multiple interaction modalities, supporting business logic and user queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If current virtual agents use large sets of tagged sentences for intent identification and entity extraction, then they can perform basic conversational tasks, but they struggle with complex business processes and lack adaptability
Solution Approach 1:
The patent replaces the mechanical system of manual sentence tagging with an automated model-based approach. Task models are declaratively defined using a task modeling language, and a natural language grammar generator automatically creates the grammar from these models, eliminating the need for manual tagging of numerous sentences while maintaining and improving capability to handle complex business processes.
Solution Approach 2:
The patent changes the fundamental parameter of how conversational agents acquire and store knowledge. Instead of storing large sets of tagged sentences, the system uses a compact task model representation that can be automatically transformed into natural language grammar, significantly reducing data requirements while improving adaptability to complex processes.
2Adaptability or versatility
If current virtual agents are trained on specific modalities, then they can function within those platforms, but they cannot provide seamless cross-modality interaction
Solution Approach 1:
The patent creates a universal task model representation that can be used across multiple interaction modalities. The same task model defined in the task modeling language can be processed to generate responses for different modalities (text, voice, etc.), allowing a single agent to function seamlessly across platforms without requiring separate modality-specific training.
3Adaptability or versatility
If current virtual agents use rigid dialog flows, then they can maintain structured interactions, but they cannot handle user interruptions or maintain context dynamically
Solution Approach 1:
The patent introduces dynamic adaptability to dialog flows through model-based processing. The task models allow the agent to dynamically adjust its behavior based on user inputs and context, enabling it to handle interruptions and maintain relevant context without requiring rigid pre-programmed dialog flows, thus improving natural human-like interaction.
Data Source
AI summary
A method of providing a conversational agent for interacting with a user may include declaratively defining a task model of a task using a task modelling language, storing the task model in a computer-readable storage medium, generating a natural language grammar based on the task model, storing the natural language grammar in the computer-readable storage medium, receiving a user input from the user, interpreting the user input with a processor based on the task model and the natural language grammar, generating an agent response to the user input with the processor based on the task model, and communicating the agent response to the user.