Speech Recognition Homonym Disambiguation via Segmented Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition and synthesis systems face challenges in accurately processing homonyms, particularly tradenames and given names, due to their similar pronunciations and varying spellings, leading to inconsistent and inefficient search results.
Innovation Solution
A language processing system that utilizes a language model incorporating dictionary entries for display forms of homonyms, enabling text normalization and inverse text normalization to select the appropriate display form of homonyms, thereby improving search accuracy and user interface presentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If text normalization techniques are used to convert symbols and digits to a smaller vocabulary set, then processing speed and efficiency are improved, but accuracy in distinguishing homonyms with similar pronunciations deteriorates
Solution Approach 1:
The patent segments the text normalization process into two distinct phases: (1) conversion to a normalized smaller vocabulary for efficient processing, and (2) subsequent inverse text normalization to restore proper display forms. This segmentation allows the system to benefit from both small-vocabulary efficiency and accurate homonym distinction by applying different processing rules at different stages.
Solution Approach 2:
The patent applies preliminary text normalization to convert spoken input to a compact vocabulary representation before processing, then applies inverse text normalization as a preliminary step before display to ensure proper homonym formatting. These preliminary actions at strategic points in the processing pipeline enable both efficiency and accuracy.
2Device complexity
If homonyms are converted to a single normalized form for processing, then vocabulary size is reduced and processing is simplified, but the ability to present accurate display forms deteriorates
Solution Approach 1:
The patent introduces an intermediary inverse text normalization step that acts as a bridge between the normalized processing form and the final display form. This intermediary process retrieves and applies the appropriate display form rules after processing, ensuring that no display form information is lost despite using a simplified vocabulary during processing.
3Measurement precision
If a larger vocabulary including all homonym variations is used, then search accuracy is improved, but processing efficiency and speed deteriorate
Solution Approach 1:
The patent implements a dynamic two-stage processing approach where the system adapts its vocabulary usage based on the processing stage: using a compact normalized vocabulary during the processing phase for efficiency, then dynamically switching to expanded display form vocabulary in the output phase for accuracy. This dynamic adaptation resolves the contradiction between speed and accuracy.
Data Source
AI summary
A language processing system may determine a display form of a spoken word by analyzing the spoken form using a language model that includes dictionary entries for display forms of homonyms. The homonyms may include trade names as well as given names and other phrases. The language processing system may receive spoken language and produce a display form of the language while displaying the proper form of the homonym. Such a system may be used in search systems where audio input is converted to a graphical display of a portion of the spoken input.