Multilingual Named Entity Speech Recognition via G2P Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional speech recognition systems struggle to accurately detect and interpret multilingual named entities, particularly when native language inputs contain foreign words or phrases, due to difficulties in distinguishing between words with the same spelling but different meanings and pronunciations across languages.
Innovation Solution
A method and system that access a list of named entities in both native and foreign languages, identify common words or phrases, and apply grapheme-to-phoneme conversion to map foreign words to native pronunciations, using a phoneme-to-phoneme converter and grapheme-to-phoneme converter to update a recognition dictionary for improved speech recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional speech recognition systems are used to detect named entities in native language inputs, then the system can process basic speech inputs, but it cannot accurately detect foreign words embedded in native language inputs due to spelling ambiguities across languages
Solution Approach 1:
The system segments the named entity detection task into multiple language-specific sub-tasks. It identifies and separates foreign words from native language inputs by comparing against multiple language dictionaries, allowing each word to be processed according to its language of origin rather than treating the entire input as a single language
Solution Approach 2:
The patent introduces an intermediary component that acts as a language identification and disambiguation layer between the speech recognition system and the named entity detection. This intermediary uses grapheme-to-phoneme conversion and phoneme comparison across languages to determine whether a word should be interpreted in the native language or as a foreign word, resolving the spelling ambiguity before final recognition
2Measurement precision
If the system applies grapheme-to-phoneme conversion and multi-language dictionary comparison to identify foreign words, then detection accuracy improves, but system complexity increases
Solution Approach 1:
The system performs preliminary grapheme-to-phoneme conversion on the input text before attempting named entity detection. By converting graphemes to phonemes early in the processing pipeline and comparing against pre-loaded multi-language dictionaries, the system resolves language ambiguity upfront, simplifying subsequent detection steps rather than adding complexity throughout the entire system
Solution Approach 2:
The patent creates a universal phoneme-based comparison mechanism that works across multiple languages simultaneously. Rather than implementing separate detection systems for each language, a single phoneme-to-phoneme mapping system handles disambiguation for all languages in the supported set, reducing overall system complexity while maintaining multi-language capability
Data Source
AI summary
Methods and systems are provided for improving speech recognition of multilingual named entities. In some embodiments, a list comprising a plurality of named entities may be accessed by a computing device. A first named entity represented in the native language may be compared with the first named entity represented in the foreign language. One or more words that appear in both the first named entity represented in the native language and the first named entity represented in the foreign language may be identified as one or more foreign words. A grapheme-to-phoneme (G2P) conversion may be applied to the one or more foreign words, wherein graphemes of the one or more foreign words are mapped to phonemes in the native language. The G2P conversion may result in a native pronunciation for each of the one or more foreign words, which are added to a recognition dictionary along with the native pronunciations.