Mixed-Initiative Dialogue System Using Global Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dialogue systems struggle to efficiently handle non-linear user behavior due to the need for pre-specifying all possible dialogue pathways, which can be computationally resource-intensive and limit storage capacity, especially in complex conversations.
Innovation Solution
A mixed-initiative design combining a state machine approach with global rules, allowing transitions based on user input and pre-defined pathways, enabling efficient modeling of non-linear conversations without requiring extensive pre-specification of dialogue pathways or states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all possible dialogue pathways are pre-specified in a state machine, then the dialogue system can handle linear conversations reliably, but the computational resource consumption and storage capacity requirements increase significantly
Solution Approach 1:
The patent segments the dialogue system into two distinct components: a state machine that handles pre-specified dialogue pathways and a global rules engine that handles non-linear user behavior. This segmentation allows the system to maintain reliability for standard conversations while avoiding the complexity of pre-specifying all possible pathways. The state machine contains only the essential dialogue flow states, while the global rules layer adds flexibility without increasing state machine complexity.
Solution Approach 2:
The patent extracts the handling of non-linear user behavior from the state machine structure and places it in a separate global rules engine. This extraction removes the burden of managing complex non-linear pathways from the state machine, reducing its complexity while maintaining the ability to handle such cases through the global rules layer that operates independently.
2Adaptability or versatility
If extensive dialogue pathways are pre-specified to handle non-linear user behavior, then conversation scenarios are handled effectively, but storage capacity is limited
Solution Approach 1:
The global rules engine serves as a universal component that can handle multiple conversational scenarios without requiring separate pre-specified pathways for each scenario. Instead of storing extensive scenario-specific pathways, the system uses a small set of global rules that can dynamically adapt to various non-linear user behaviors, significantly reducing storage requirements while maintaining versatility.
Solution Approach 2:
The system changes the parameter of dialogue pathway representation from extensive pre-specified state transitions to compact global rules with conditions and actions. This parameter change allows the system to represent complex conversational scenarios using minimal storage, as the global rules engine can generate appropriate responses dynamically based on the current dialogue state and user input, rather than storing every possible pathway.
3Stability of the object's composition
If a state machine approach is used to define dialogue pathways, then structured dialogue flow is achieved, but the system requires extensive pre-specification of states and transitions
Solution Approach 1:
The patent segments the dialogue management functionality between the state machine (providing structured flow) and the global rules engine (providing flexibility). The state machine contains only the essential structured pathways needed for stable dialogue flow, while the global rules layer handles the complexity of non-linear behaviors, allowing the state machine to remain simple while maintaining structural stability.
Solution Approach 2:
The global rules engine acts as an intermediary between the simple state machine structure and the complex non-linear user behavior. It mediates by intercepting user inputs that don't fit the pre-specified pathways and routing them through appropriate global rules, thereby maintaining the stability of the state machine structure while handling complexity without requiring extensive state specification.
Data Source
Figure 1
Figure 2(a)
Figure 2(b)
AI summary
There is provided a dialogue system comprising: an input for obtaining an input signal relating to speech or text input provided by a user; an output for outputting speech or text information specified by a determined dialogue act; and a processor configured to: determine dialogue information from the input signal; determine the dialogue act based on the determined dialogue information, wherein determining the dialogue act comprises: selecting a next state from a plurality of states, wherein each of the plurality of states comprises information specifying a dialogue act and transition information specifying a transition to another state, the transitions defining one or more dialogue pathways, wherein selecting the next state comprises selecting a state which is specified by the transition information in a current state or selecting a state which is specified by a rule in a first set of one or more rules.