Dynamic State Machine Generation for Conversational Flow
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated response systems for natural language queries are resource-intensive and inefficient, particularly when handling multiple users and vast conversation paths, leading to excessive computing resource utilization.
Innovation Solution
Dynamically generate state machines based on intent identifiers using conversational flow templates, which are tailored to specific queries, reducing resource usage by focusing on relevant conversation paths and deleting state machines after use.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If rules or logic are used to determine automated responses to natural language queries, then the system can handle diverse conversation paths, but the computing memory and processing resources required to load and execute such logic at runtime become prohibitively large
Solution Approach 1:
The patent segments the conversational logic into discrete states and transitions, where each state represents a specific point in the conversation flow and transitions define valid paths between states. This segmentation allows the system to represent diverse conversation paths through a manageable graph structure rather than loading all possible rules into memory simultaneously.
Solution Approach 2:
The patent implements dynamic state machine generation where the conversation flow model is constructed at runtime based on the specific query and user context, rather than loading a static complete logic graph. The state machine adapts its structure dynamically, creating only the necessary paths for the current conversation scenario, thereby reducing resource consumption while maintaining adaptability.
2Adaptability or versatility
If all possible conversation paths are tracked for multiple users, then the system can provide personalized responses, but the computing resource utilization becomes prohibitively large
Solution Approach 1:
The patent applies local quality by customizing the state machine for each user based on their specific context, history, and preferences rather than maintaining a single monolithic conversation model for all users. Each user receives a tailored state machine that includes only the relevant states and transitions for their specific interaction pattern, reducing overall computing resource utilization while preserving personalization capability.
Solution Approach 2:
The system dynamically generates user-specific state machines at runtime based on individual user context rather than pre-loading complete conversation models for all users. This dynamic approach allows the system to allocate computing resources proportionally to actual usage needs, providing personalized responses only when necessary and reducing overall energy consumption.
3Adaptability or versatility
If a comprehensive state machine is generated for all possible conversation paths, then the system can handle any query, but the computing resources required to generate and execute the state machine become excessive
Solution Approach 1:
The patent applies partial action by generating state machines that include only the necessary states and transitions relevant to the current query and user context, rather than creating comprehensive state machines for all possible conversations. This selective generation approach maintains sufficient query handling capability for the immediate task while significantly reducing the complexity and resource requirements of the state machine execution.
Solution Approach 2:
The system dynamically determines the scope and structure of the state machine based on the specific query and user profile, generating only the necessary conversational paths for the current interaction. This dynamic state machine generation ensures the system handles the current query effectively while avoiding the excessive complexity of pre-generating all possible conversation paths.
Data Source
AI summary
Aspects of the present disclosure provide techniques for dynamic state machine based conversational flow execution in a software application. Embodiments include receiving a natural language query via a user interface and using a classification machine learning model to determine an intent identifier based on the natural language query. Embodiments include selecting, based on the intent identifier, a conversational flow template that specifies conversational logic via blocks and edges. Embodiments include dynamically generating a state machine comprising an initial state corresponding to a point within the selected conversational flow template that is associated with the intent identifier and one or more additional states and conditions generated based on a subset of the blocks and the edges of the conversational flow template. Embodiments include executing the dynamically generated state machine in order to automatically generate a response to the natural language query and providing the response via the user interface.