Google LLC (20240290323). Large-Scale Language Model Data Selection for Rare-Word Speech Recognition simplified abstract
Contents
Large-Scale Language Model Data Selection for Rare-Word Speech Recognition
Organization Name
Inventor(s)
Wenqian Ronny Huang of Mountain View CA (US)
Tara N. Sainath of Jersey City NJ (US)
Large-Scale Language Model Data Selection for Rare-Word Speech Recognition - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240290323 titled 'Large-Scale Language Model Data Selection for Rare-Word Speech Recognition
- Simplified Explanation:**
The patent application describes a method for training a language model for rare-word speech recognition by filtering out rare words from training text samples and training utterances.
- Key Features and Innovation:**
- Obtaining training text samples and training utterances for speech recognition model training.
- Applying rare word filtering to identify rare-word training text samples.
- Training the language model on transcriptions from training utterances and rare-word training text samples.
- Potential Applications:**
This technology can be used in speech recognition systems, virtual assistants, transcription services, and language learning applications.
- Problems Solved:**
This technology addresses the challenge of accurately recognizing and transcribing rare words in speech, improving the overall performance of language models.
- Benefits:**
- Enhanced accuracy in recognizing rare words in speech.
- Improved performance of language models in speech recognition tasks.
- Better transcription quality for audio data containing rare words.
- Commercial Applications:**
- Title: "Advanced Speech Recognition Technology for Improved Transcription Services"
- This technology can be commercially applied in transcription services, virtual assistant devices, language learning platforms, and customer service automation systems.
- Questions about the Technology:**
1. How does this technology improve the accuracy of speech recognition for rare words?
- This technology improves accuracy by training the language model on a subset of rare-word training text samples.
2. What are the potential applications of this technology beyond speech recognition?
- This technology can also be applied in language learning platforms and virtual assistant devices for improved performance.
Original Abstract Submitted
a method of training a language model for rare-word speech recognition includes obtaining a set of training text samples, and obtaining a set of training utterances used for training a speech recognition model. each training utterance in the plurality of training utterances includes audio data corresponding to an utterance and a corresponding transcription of the utterance. the method also includes applying rare word filtering on the set of training text samples to identify a subset of rare-word training text samples that include words that do not appear in the transcriptions from the set of training utterances or appear in the transcriptions from the set of training utterances less than a threshold number of times. the method further includes training the external language model on the transcriptions from the set of training utterances and the identified subset of rare-word training text samples.
