Neural Utterance Generation for Virtual Assistant Intent Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual assistants struggle to accurately recognize and handle user requests due to the endless variations in how users describe the same query, requiring developers to manually define numerous phrases, which is labor-intensive and inefficient.
Innovation Solution
Employing a neural network language model trained via unsupervised learning to generate a comprehensive set of sample utterance sentences that can invoke specific intents, using finetuned natural language generators and classifiers to improve accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If developers manually define all possible ways users can describe a request, then the virtual assistant can accurately recognize standard queries, but the device complexity and developer effort increase significantly
Solution Approach 1:
The system uses unsupervised learning to automatically generate utterance sentences without requiring manual definition by developers. The neural network model learns from existing data and autonomously creates the phrases needed for intent recognition, eliminating the need for developers to manually define every possible user query variation.
Solution Approach 2:
The patent replaces the mechanical process of manual phrase definition with an automated neural network-based system. Instead of requiring developers to systematically create and manage extensive phrase lists, the system uses machine learning algorithms to generate and manage utterance sentences dynamically.
2Reliability
If developers manually write and verify every possible utterance sentence, then the virtual assistant can handle standard queries well, but the time required for configuration increases
Solution Approach 1:
The system performs preliminary generation of utterance sentences automatically using unsupervised learning before the virtual assistant needs to handle user queries. By pre-generating a comprehensive set of possible phrases through neural network models, the system eliminates the need for developers to manually create and verify each phrase during configuration.
Solution Approach 2:
The neural network model autonomously generates and verifies utterance sentences without requiring developer intervention. The system self-services the configuration process by automatically creating the phrase database that enables accurate intent recognition, significantly reducing the time developers would otherwise need to spend on manual creation and verification.
3Device complexity
If the virtual assistant uses a limited set of predefined phrases, then the system remains simple to manage, but it fails to recognize variations in user language
Solution Approach 1:
The system changes the parameter of phrase generation from static, manually-defined lists to dynamic, algorithmically-generated sentences. By using unsupervised learning models that can adapt to different language patterns and generate varied utterance sentences, the system maintains simplicity in management while significantly improving its ability to handle language variations.
Solution Approach 2:
The patent replaces the static mechanical approach of predefined phrase lists with a dynamic neural network-based generation system. This substitution enables the virtual assistant to automatically adapt to various user language patterns without requiring manual updates to predefined phrases, thereby improving versatility while keeping the system simple to manage.
Data Source
AI summary
Methods and systems for automatically generating sample phrases or sentences that a user can say to invoke a set of defined actions performed by a virtual assistant are disclosed. By enabling finetuned general-purpose natural language models, the system can generate potential and accurate utterance sentences based on extracted keywords or the input utterance sentence. Furthermore, domain-specific datasets can be used to train the pre-trained, general-purpose natural language models via unsupervised learning. These generated sentences can improve the efficiency of configuring a virtual assistant. The system can further optimize the effectiveness of a virtual assistant in understanding the user, which can enhance the user experience of communicating with it.