Back to results
US 12,718,798 B2Granted
Electronic device and method for creating customized language model
AssigneeSamsung Electronics Co., Ltd.
InventorsGunu Jho, Hoseon Shin
Life of this patent
Estimating the term…
Abstract
An example electronic device may include a memory configured to store instructions and a processor electrically connected to the memory and configured to execute the instructions. When the instructions are executed by the processor, the processor may be configured to create an automatic speech recognition (ASR) language model including information about a plurality of candidate transliterations for a variously utterable text, based on a context of a user indicating a situation of the user, a basic language model, or a customized language model and update the customized language model in response to an utterance of the user matching one of the plurality of candidate transliterations.
Claims14
An electronic device comprising:
- memory configured to store instructions;
- and
- at least one processor, comprising processing circuitry, electrically connected to the memory and configured, individually or collectively, to execute the instructions and to control the electronic device to:
- train a transliteration model based on training data comprising a corpus and a transliteration of the corpus by inputting the corpus to a pronunciation sequence prediction model to obtain a pronunciation of the corpus and inputting the pronunciation of the corpus to a phoneme conversion model to obtain a grapheme converted into a user-specified language;
- create an automatic speech recognition (ASR) language model comprising information about a plurality of candidate transliterations obtained from the transliteration model for a same text utterable in various different ways, based on a context of a user indicating a situation of the user;
- receive, via a microphone, an utterance of the user;
- update a language model customized to language characteristics of the user in response to the utterance of the user matching one of the plurality of candidate transliterations;
- and
- provide, via a speaker, a response corresponding to the utterance of the user in which the variously utterable text is uttered in a manner customized for the user, based on the updated customized language model.
6 more elements
An electronic device comprising:
- memory configured to store instructions;
- and
- at least one processor, comprising processing circuitry, electrically connected to the memory and configured, individually or collectively, to execute the instructions and to control the electronic device to:
- train a transliteration model based on training data comprising a corpus and a transliteration of the corpus by inputting the corpus to a pronunciation sequence prediction model to obtain a pronunciation of the corpus and inputting the pronunciation of the corpus to a phoneme conversion model to obtain a grapheme converted into a user-specified language;
- receive, via a microphone, an utterance of a user in which a text comprising a first language is expressed in a second language;
- recognize the utterance and provide a response, based on an automatic speech recognition (ASR) language model created based on a context of the user indicating a situation of the user and comprising information about a plurality of candidate transliterations transliterated into the second language for the text using the transliteration model;
- convert the utterance of the user into text data;
- match the text data with the plurality of candidate transliterations;
- update a customized language model by determining a matched candidate transliteration as a correct answer for the text comprising the first language when the text data matches one of the plurality of candidate transliterations;
- and
- provide a response of uttering the text comprising the first language in a same manner that the correct answer utters the text.
8 more elements
A method of operating an electronic device, the method comprising:
- training a transliteration model based on training data comprising a corpus and a transliteration of the corpus by inputting the corpus to a pronunciation sequence prediction model to obtain a pronunciation of the corpus and inputting the pronunciation of the corpus to a phoneme conversion model to obtain a grapheme converted into a user-specified language;
- creating an automatic speech recognition (ASR) language model comprising information about a plurality of candidate transliterations obtained from the transliteration model for a same text utterable in various different ways, based on a context of a user indicating a situation of the user;
- receive, via a microphone, an utterance of the user;
- updating a customized language model customized to language characteristics of the user in response to an utterance of the user matching one of the plurality of candidate transliterations;
- and
- providing, via a speaker, a response corresponding to the utterance of the user in which the variously utterable text is uttered in a manner customized for the user, based on the updated customized language model.
3 more elements
Dependencies are inferred from the claim text.