Virtual Assistant Request Recognition via Salient Dialog Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual assistants struggle to recognize and respond to new and evolving user requests due to outdated training data and unfamiliar terminology, leading to frustration for users who require assistance with diverse device functionalities.
Innovation Solution
The system compares user requests to known templates, generates candidate templates from unrecognized requests, and trains the virtual assistant using salient templates identified through frequent usage, as well as anticipated future data to populate request templates, enabling the assistant to adapt and recognize future user needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a virtual assistant uses predetermined request templates for recognition, then it can reliably respond to known requests, but it cannot recognize or respond to new requests with evolving terminology
Solution Approach 1:
The system proactively collects and analyzes unrecognized user requests in advance, identifying patterns and generating candidate templates before they become common enough to be automatically recognized. This preliminary action allows the virtual assistant to adapt to new terminology and requests before they would otherwise be missed, resolving the contradiction between reliable predetermined recognition and adaptability to new requests.
Solution Approach 2:
The system implements a feedback loop where unrecognized requests are captured, analyzed for patterns, and used to generate new candidate templates that are then added to the recognition system. This continuous feedback mechanism enables the virtual assistant to learn from user interactions and improve its request recognition capabilities over time, balancing reliability with adaptability.
2Adaptability or versatility
If the virtual assistant collects and processes all user requests to improve recognition, then it can adapt to new requests, but the system complexity and processing overhead increase
Solution Approach 1:
The system applies different processing quality to different types of requests: unrecognized requests receive intensive analysis and candidate template generation, while recognized requests follow standard processing paths. This local quality approach allows the system to maintain high adaptability for new requests while avoiding unnecessary complexity for routine requests, resolving the contradiction between adaptability and system complexity.
Solution Approach 2:
The system changes the threshold parameter for template salience over time, starting with lower thresholds to quickly adapt to new requests and gradually increasing thresholds to reduce processing overhead as the system becomes more familiar with common request patterns. This dynamic parameter adjustment resolves the contradiction between adaptability and complexity by adapting the processing intensity to the current state of the system.
3Productivity
If the system uses frequent usage counts to identify salient templates, then it can prioritize important requests, but it may overlook rare but important requests
Solution Approach 1:
The system applies partial action by focusing intensive analysis on candidate templates that meet a minimum frequency threshold, while still maintaining a mechanism to occasionally process and evaluate rare requests. This approach prioritizes processing efficiency for common patterns while preserving the ability to discover and adapt to rare but important requests, resolving the contradiction between productivity and recognition coverage.
Data Source
AI summary
Systems and processes are disclosed for virtual assistant request recognition using live usage data and data relating to future events. User requests that are received but not recognized can be used to generate candidate request templates. A count can be associated with each candidate request template and can be incremented each time a matching candidate request template is received. When a count reaches a threshold level, the corresponding candidate request template can be used to train a virtual assistant to recognize and respond to similar user requests in the future. In addition, data relating to future events can be mined to extract relevant information that can be used to populate both recognized user request templates and candidate user request templates. Populated user request templates (e.g., whole expected utterances) can then be used to recognize user requests and disambiguate user intent as future events become relevant.