Meta Intent Model for Conversational Automation Agents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conversational automation agents require redundant user interactions and computational resources when handling complex tasks that involve multiple intents, as they typically need separate dialogues for each step of the task, leading to cumbersome and inefficient communication.
Innovation Solution
A meta intent model is generated using a conversational similarity metric to recognize related intents and slots across multiple conversations, allowing for the reuse of common slots and reducing the number of user instructions needed to specify complex tasks, thereby streamlining interactions and resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If separate dialogs are used for each task step, then task completion is achieved, but user interaction becomes lengthy and cumbersome
Solution Approach 1:
The patent merges multiple separate task dialogs into a single unified dialog by introducing a parent intent that coordinates multiple child intents. This allows the automation agent to handle complex multi-step tasks in one continuous conversation rather than requiring separate dialogs for each step, thereby reducing user interaction time while maintaining task completion efficiency
Solution Approach 2:
The parent intent structure provides a universal framework that can accommodate multiple different child intents and tasks within a single dialog. This multi-functional approach allows the same dialog structure to handle various complex tasks by dynamically activating relevant child intents, improving productivity without requiring separate specialized dialogs for each task type
2Productivity
If separate dialogs are used for each task step, then task completion is achieved, but computational resources are over-used
Solution Approach 1:
By merging multiple task handling processes into a single unified dialog managed by the parent intent, the system reduces the total computational overhead. Instead of initializing and managing multiple separate dialog contexts, the parent intent consolidates resource usage while still coordinating all necessary child intents to complete the complex task efficiently
3Reliability
If multiple separate dialogs are conducted for complex tasks, then each task step is completed, but redundant communications occur
Solution Approach 1:
The parent intent merges multiple child intent executions into a single dialog context, allowing the system to maintain reliable task step completion while eliminating redundant communications. Information exchanged in one part of the dialog can be reused for subsequent steps, preventing the need to re-request the same information in separate dialogs and reducing communication redundancy
Data Source
AI summary
Automating generalization or personalization of conversational automation agents includes receiving, by computer hardware, a plurality of input conversations. The input conversations include, or are formed of, a plurality of utterances. A plurality of intents and slots are determined from the input conversations by processing the plurality of input conversations through a first classifier. A plurality of generalized intents are generated by performing entity recognition on the plurality of intents and slots using an entity recognizer. The entity recognizer is configured to apply a knowledge graph to the plurality of intents and slots. Slots of the plurality of input conversations as classified are masked to generate masked utterances. Conversational data, which includes the masked utterances and the plurality of generalized intents, are encoded as a plurality of feature vectors. A meta intent model is generated by processing the plurality of feature vectors through a second classifier using a conversation similarity metric.