Natural Assistant Interaction with In-Stream Trigger Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual assistants require a trigger phrase at the beginning of every user utterance, making human-machine interaction cumbersome and less natural.
Innovation Solution
A virtual assistant system that allows a trigger word or phrase to be placed anywhere within an audio stream, eliminating the need for a trigger phrase at the beginning, and uses a false-trigger mitigator to improve accuracy in determining user intent.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a trigger phrase is required at the beginning of every user utterance, then the virtual assistant can reliably identify user intent, but the interaction becomes cumbersome and less natural
Solution Approach 1:
The patent extracts the trigger phrase detection from the beginning-of-utterance position and allows it to occur anywhere within the audio stream. The system separates the trigger phrase detection function from the structural position requirement, enabling the trigger to be embedded within natural speech rather than requiring it to lead the utterance.
Solution Approach 2:
The system dynamically adjusts its processing based on whether a trigger phrase is detected anywhere in the audio stream. When a trigger is present, the system processes the entire utterance for intent recognition; when no trigger is present, it may disregard the input or process it differently. This dynamic behavior allows flexible interaction patterns.
2Ease of operation
If the virtual assistant processes every audio stream as a potential user request, then no trigger phrase is needed, but false triggers increase and processing efficiency decreases
Solution Approach 1:
The system performs preliminary detection of trigger phrases within the audio stream before proceeding to full intent recognition processing. This preliminary action filters out audio streams that do not contain valid triggers, preventing false triggers from consuming processing resources while maintaining the ability to respond to legitimate user requests.
Solution Approach 2:
The trigger phrase acts as an intermediary element that mediates between the user's natural speech and the virtual assistant's intent recognition system. By requiring this intermediary signal, the system can distinguish between genuine user requests and incidental audio input, improving reliability without completely eliminating natural speech patterns.
3Measurement precision
If the virtual assistant generates candidate intents for every audio stream, then user intent can be accurately determined, but power consumption increases
Solution Approach 1:
The system applies partial processing by generating candidate intents only for audio streams that contain detected trigger phrases. Instead of performing full intent recognition on all audio input, it selectively applies the computationally intensive candidate generation process only when warranted by the presence of a trigger, thereby reducing overall power consumption while maintaining accuracy for legitimate requests.
Data Source
AI summary
Systems and processes for operating a virtual assistant to provide natural assistant interaction are provided. In accordance with one or more examples, a method includes, at an electronic device with one or more processors and memory: receiving a first audio stream including one or more utterances; determining whether the first audio stream includes a lexical trigger; generating one or more candidate text representations of the one or more utterances; determining whether at least one candidate text representation of the one or more candidate text representations is to be disregarded by the virtual assistant. If at least one candidate text representation is to be disregarded, one or more candidate intents are generated based on candidate text representations of the one or more candidate text representations other than the to be disregarded at least one candidate text representation.