Voice Conversation Continuity Detection via Contextual Transcription Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Natural language processing systems face challenges in determining whether a user's subsequent spoken requests are intended for the device without the use of a hotword, leading to ambiguity and inefficient user interaction.
Innovation Solution
The system analyzes the content of follow-up utterances in conjunction with other signals such as previous utterances, responses, screen content, location, and time to determine if a user is continuing a conversation with the device, allowing it to route the transcription appropriately for processing or bypassing it if not intended for the device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the device requires a hotword before each utterance to determine user intent, then the device can reliably identify when speech is directed to it, but the user interaction becomes awkward and inefficient
Solution Approach 1:
The system performs preliminary analysis of the first utterance and device response to establish context before the second utterance occurs. By pre-processing the conversation history and identifying relevant terms, the system prepares the groundwork for quickly determining whether subsequent speech is directed to the device, eliminating the need for repeated hotwords and improving interaction efficiency while maintaining reliable detection
Solution Approach 2:
The system uses the device's response to the first utterance as feedback to help determine whether the second utterance is directed to the device. By analyzing whether terms in the second utterance match or are related to terms in the first utterance and the device response, the system creates a feedback loop that maintains detection accuracy without requiring additional hotwords, thus resolving the contradiction between reliability and ease of operation
2Ease of operation
If the device allows follow-up utterances without hotwords, then user interaction becomes more natural and efficient, but the device cannot reliably determine whether the speech is intended for it
Solution Approach 1:
The system introduces the transcription analysis as an intermediary mechanism between the second utterance and the determination of user intent. By transcribing the second utterance and comparing its terms against the first utterance and device response, the system creates an intermediate analysis layer that enables reliable detection without requiring hotwords, thus allowing natural interaction while maintaining accuracy
Solution Approach 2:
The system performs preliminary transcription and term extraction from the first utterance and device response before analyzing the second utterance. This pre-processing creates a reference framework that enables the system to quickly and reliably determine whether the second utterance is directed to the device, making the interaction feel natural while maintaining detection accuracy through advance preparation
3Reliability
If the device analyzes transcription content and context to determine user intent, then it can accurately identify follow-up utterances without hotwords, but the processing complexity increases
Solution Approach 1:
The system segments the analysis into distinct components: extracting terms from the first utterance, identifying terms in the device response, and comparing these against terms in the second utterance. By dividing the complex analysis task into manageable segments, the system achieves accurate detection of user intent while keeping processing complexity manageable through structured, modular analysis steps
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for detecting a continued conversation are disclosed. In one aspect, a method includes the actions of receiving first audio data of a first utterance. The actions further include obtaining a first transcription of the first utterance. The actions further include receiving second audio data of a second utterance. The actions further include obtaining a second transcription of the second utterance. The actions further include determining whether the second utterance includes a query directed to a query processing system based on analysis of the second transcription and the first transcription or a response to the first query. The actions further include configuring the data routing component to provide the second transcription of the second utterance to the query processing system as a second query or bypass routing the second transcription.