Deep Multi-View Clustering for Automated Goal Discovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated dialog systems struggle to discover user goals in complex conversations, especially in domains like customer support and healthcare, as they rely on supervised clustering of initial utterances, which is time-consuming and inefficient, and fail to account for goal evolution over time.
Innovation Solution
A system using deep multi-view clustering with neural networks that partitions conversation datasets into two views, alternates between generating and clustering vector representations, and updates encoders to improve agreement between cluster assignments, allowing for unsupervised discovery of user goals throughout the conversation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised clustering of user first utterances is used to discover goals, then goal discovery can be achieved, but it requires a lot of human time in tagging the dataset
Solution Approach 1:
The system performs unsupervised clustering that automatically discovers goals from conversation data without requiring human annotation. The neural network encoders and clustering algorithm work autonomously to identify goal patterns, eliminating the need for manual dataset tagging while maintaining goal discovery capability
Solution Approach 2:
The patent replaces the manual mechanical process of human tagging with an automated computational system using neural networks and clustering algorithms. The deep multi-view clustering framework substitutes human labor with machine-based pattern recognition, achieving both time efficiency and accurate goal discovery
2Loss of time
If unsupervised clustering is used to discover goals, then human time is reduced, but it has proved challenging due to semantic variations in conversations
Solution Approach 1:
The patent segments the conversation data into two distinct views: user utterances and system responses. This segmentation allows the system to capture semantic information from multiple perspectives, improving clustering reliability by considering both sides of the interaction rather than relying on a single view
Solution Approach 2:
The patent introduces a multi-view dimension to the clustering problem by creating separate encoders for user and system perspectives. This dimensional expansion allows the system to handle semantic variations more effectively by mapping conversations into a richer feature space that captures relationships across different interaction dimensions
3Device complexity
If only user query/request is used to express goal, then goal discovery is simplified, but goals are not fully captured as user goal is expressed in the rest of the conversation too
Solution Approach 1:
The patent creates encoders that are universal to both user and system perspectives, allowing the same clustering framework to process information from both sides of the conversation. This multi-functional approach ensures complete goal capture while maintaining process coherence
Solution Approach 2:
The patent merges information from user utterances and system responses into a unified clustering process. By combining both views through the deep multi-view clustering framework, the system achieves complete goal information capture while managing complexity through integrated processing
4Productivity
If conversations are clustered by initial utterances only, then processing is faster, but goal evolution over time is not captured
Solution Approach 1:
The patent applies clustering to the entire conversation sequence rather than just initial utterances, maintaining continuous analysis throughout the interaction. This allows the system to track goal evolution over time while preserving processing efficiency through automated neural network-based clustering
Solution Approach 2:
The patent employs dynamic clustering that can adapt to changing goals throughout the conversation. The deep multi-view clustering framework processes the full conversation history, allowing cluster assignments to reflect goal evolution and temporal changes in user intent
Data Source
AI summary
The present disclosure describes a system, method, and computer program for automatically discovering goals from conversations using neural networks and deep multi-view clustering. A dataset of conversations is partitioned into two views. Vector representations of each view are then generated and clustered in an alternating fashion between views for a number of iterations (i.e., the system alternates between views in generating and clustering vector representations of a view). A first neural network encoder generates the vector representations for the first view, and a second neural network encoder generates the vector representations for the second view. With each semi-iteration, cluster assignments from one view are used to update the encoder for the other view, thus encouraging the two neural network encoders to yield similar cluster assignments. After all the iterations are complete, a user interface enables a user to label each first-view cluster with a goal, where a subset of example utterances is displayed for each cluster.