Intent-Based Conversational Knowledge Graph for Spoken Language Understanding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Spoken Language Understanding (SLU) systems face challenges in accurately extracting user intent from input phrases, particularly due to the need for large amounts of manually labeled training data and limitations in handling new topics or intents, leading to compromised accuracy and limited capabilities.
Innovation Solution
The implementation of a scale-free network structured conversational knowledge graph that uses Universal Sentence Embedding (USE) models and graph query techniques to automatically process intent-based spoken language, allowing for the expansion of topic/intent capabilities and improved accuracy with reduced training data requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised learning with manually labeled training data is used to extract user intent, then the system can achieve intent classification capability, but the amount of training data required becomes large and the system cannot handle new topics beyond the training scope
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary structure that stores semantic relationships between intents, entities, and attributes. This knowledge graph serves as a mediator between the input phrase and the intent classification process, enabling the system to leverage structured semantic knowledge rather than relying solely on large volumes of labeled training data. The knowledge graph allows the system to generalize to new topics by utilizing pre-established semantic relationships.
2Adaptability or versatility
If pattern matching with pre-built lexical and syntactic patterns is used to extract intent, then the system can process known patterns, but it cannot automatically cover new topics or intents and accuracy is compromised
Solution Approach 1:
The patent implements a dynamic pattern generation mechanism that automatically creates lexical and syntactic patterns from the knowledge graph based on the input phrase. Instead of using static pre-built patterns, the system dynamically adapts the pattern matching process to the specific input, allowing it to handle new topics and intents while maintaining accuracy through the structured semantic guidance provided by the knowledge graph.
3Ease of operation
If a conventional SLU system is designed to extract meaning from user utterances, then it can process language input, but it requires vast and broad applications with compromised accuracy when automation is involved
Solution Approach 1:
The patent segments the intent extraction process into distinct stages: input phrase processing, knowledge graph query execution, and result aggregation. By dividing the complex task of intent extraction into manageable segments, each stage can be optimized independently. The knowledge graph query stage specifically segments the problem of handling diverse language inputs by routing them through structured semantic relationships, thereby improving overall accuracy without sacrificing ease of operation.
Data Source
AI summary
Described are systems, methods, apparatuses, and computer program product embodiments for automatically processing intent-based spoken language for SLU. The disclosed solution uses a scale-free network structured conversational knowledge graph that stores nodes representative of actions, objects, and intent names and edges representative of relationships between the nodes. For all phrases (including a sentence) from the same intent, the system calculates a mean feature vector using a Universal Sentence Embedding (USE) model as a feature element. The system also employs a multi-step intent detection strategy. A graph query technique may be used to match all potential intent nodes from the trained knowledge graph. The system may compute a covariance matrix between the feature element of an input phrase and feature elements of all potential intents. The major component of the covariance matrix along with the maximum covariance may be used to determine the final intent.