Speech Intent Detection via Similarity Thresholding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition systems fail to accurately interpret user intentions due to inaccuracies in pronunciation and command confusion, leading to misrecognized words and incomplete responses, as they do not adequately consider previous user speech when analyzing new input.
Innovation Solution
An electronic apparatus and method that compares the similarity between previous and new user speech to provide a response, using a processor to identify intention and target words, acquire additional speech if necessary, and generate a user interface to guide the user in rephrasing, ensuring a threshold similarity is met for accurate response generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the electronic apparatus analyzes only new user speech to determine user intention, then the response can be provided quickly, but the accuracy of intention detection deteriorates when new speech is incomplete or supplementary
Solution Approach 1:
The system pre-establishes a database of stored words and their corresponding intentions before the user speaks. When new speech is received, the system compares it against this pre-existing database to quickly identify matches, avoiding the need to analyze all possible word combinations from scratch and enabling rapid accurate response even when speech is incomplete
Solution Approach 2:
The system provides feedback by comparing the new speech with stored words and notifying the user when the similarity is insufficient. This feedback mechanism allows the user to clarify their intention with additional speech, and the system continuously updates its understanding until sufficient accuracy is achieved, resolving the contradiction between quick response and accurate detection
2Reliability
If the electronic apparatus waits for complete user speech before providing response, then the accuracy of intention detection is improved, but the productivity of the system deteriorates
Solution Approach 1:
The system performs partial action by analyzing only the portion of new speech that achieves sufficient similarity with stored words. Instead of requiring complete speech before responding, the system can provide responses based on partial matches that meet a threshold criteria, thereby improving productivity while maintaining acceptable accuracy
Solution Approach 2:
The system changes the parameter of similarity threshold dynamically. When the similarity between new speech and stored words meets or exceeds the threshold, the system transitions from waiting mode to response provision mode. This parameter-based approach allows the system to balance between accuracy and productivity by adjusting when sufficient information has been received
3Measurement precision
If the electronic apparatus uses speech recognition without considering previous speech context, then the system complexity is reduced, but the measurement precision of user intention deteriorates
Solution Approach 1:
The system segments the speech analysis process into distinct components: storing words and intentions in a database, receiving new speech, comparing new speech with stored words, and determining similarity. This segmentation allows the system to achieve precise intention analysis by combining simple, well-defined operations rather than using a single complex analysis mechanism
Data Source
AI summary
An electronic apparatus includes: a memory storing one or more commands; and a processor connected to the memory and configured to control the electronic apparatus, wherein the processor is configured, by executing the one or more instructions, to: identify a first intention word and a first target word from first speech, acquire second speech received after the first speech based on at least one of the identified first intention word or the identified first target word not matching a word stored in the memory, acquire a similarity between the first speech and the second speech, and acquire response information based on the first speech and the second speech based on the similarity being a threshold value or more.