Speech Recognition Using Text Message Context
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatic speech recognition systems face challenges in accurately interpreting speech responses, especially in environments with colloquialisms and abbreviations, such as those found in text messaging, due to the complexity of conversational and emotional contexts.
Innovation Solution
A method involving a speech recognition system that processes speech signals by receiving utterances, pre-processing them to extract acoustic data, identifying conversational context-specific acoustic models, and decoding the data to produce hypotheses, with the option to communicate uncertain data to a server for further processing, while also considering emotional contexts to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional speech recognition systems are used, then the system structure is simple, but the recognition accuracy deteriorates when dealing with colloquialisms and abbreviations in text messaging contexts
Solution Approach 1:
The system performs preliminary classification of the input context (text message, email, general conversation) before selecting the acoustic model. This preliminary action allows the system to prepare the appropriate model in advance, improving recognition accuracy for colloquialisms and abbreviations without requiring all models to be loaded simultaneously, thus managing complexity
Solution Approach 2:
The system changes the parameter of acoustic model selection based on the detected conversational context. Different acoustic models are selected for different contexts (e.g., text message context uses models trained on abbreviations and colloquialisms, while general conversation uses standard models), thereby improving accuracy for specific contexts without permanently increasing overall system complexity
2Measurement precision
If multiple context-specific acoustic models are maintained, then speech recognition accuracy for specific contexts improves, but memory requirements and processing overhead increase
Solution Approach 1:
The system performs preliminary context classification to determine which acoustic model is needed before processing the speech signal. This allows the system to load only the necessary model into memory at any given time rather than maintaining all models simultaneously, reducing memory requirements while preserving context-specific accuracy
Solution Approach 2:
The acoustic model selection is dynamic rather than static. The system can switch between different acoustic models based on the detected conversational context, allowing efficient use of memory resources by loading only the required model for the current context while maintaining the capability to switch to other models when needed
3Measurement precision
If the system processes all speech inputs with high-level context analysis, then recognition accuracy improves, but processing time increases
Solution Approach 1:
The system performs preliminary analysis of the input to determine the conversational context (text message, email, general conversation) before applying the full context-specific acoustic model. This preliminary action enables quick routing to the appropriate model, reducing processing time while maintaining accuracy by avoiding unnecessary analysis for simple cases
Solution Approach 2:
The system applies partial context analysis based on the input characteristics. For clear, unambiguous inputs, a simpler processing path is used. For ambiguous or context-dependent inputs, the full context-specific acoustic model analysis is applied. This selective approach maintains high accuracy when needed while reducing processing time for straightforward cases
Data Source
AI summary
A method of automatic speech recognition. An utterance is received from a user in reply to a text message, via a microphone that converts the reply utterance into a speech signal. The speech signal is processed using at least one processor to extract acoustic data from the speech signal. An acoustic model is identified from a plurality of acoustic models to decode the acoustic data, and using a conversational context associated with the text message. The acoustic data is decoded using the identified acoustic model to produce a plurality of hypotheses for the reply utterance.