Samsung Electronics Co., Ltd.

Minyoung Mun

Ki Soo Kwon

Sanghyun Yoo

Hodong Lee

Inchul Song

YoungSang Choi

Hwidong Na

Sang Hyun Yoo

Young Sang Choi

Seijoon Kim

Youngki Park

A method generates an anchor image embedding vector for an anchor image using an image representation model, determine first similarities between the anchor image and negative samples of the anchor image using first image embedding vectors for the negative samples and the generated anchor image embedding vector, determine second similarities between the anchor image and positive samples of the anchor image using second image embedding vectors for the positive samples and the generated anchor image embedding vector, obtain one of a vector corresponding to a label of the anchor image and third similarities between the label of the anchor image and labels of the negative samples, determine a loss value for the anchor image based on the determined first similarities, and the determined second similarities, and one of the obtained third similarities and a fourth similarity.

Method of training image representation model

A method and apparatus for personalizing a speech recognition model is disclosed. The apparatus may obtain feedback data that is a result of recognizing a first speech input of a user using a trained speech recognition model, determine whether to update the speech recognition model based on the obtained feedback data, and selectively update, dependent on the determining, the speech recognition model based on the feedback data.

Method and apparatus with a personalized speech recognition model

Disclosed is a speech recognition method and apparatus, wherein the apparatus acquires first outputs from sub-models in a recognition model based on a speech signal, acquires a second output including values corresponding to the sub-models from a classification model based on the speech signal, and recognizes the speech signal based on the first outputs and the second output.

Speech recognition method and apparatus

A translation method and apparatus may respectively perform or include: using one or more processors, plural different translation processes, in parallel, for a source sentence in a first language, including encoding, to generate respective feature vectors, the source sentence in each of two or more translation processes of the plural translation processes or the source sentence and a variation of the source sentence in respective translation processes of the plural translation processes, and decoding each of the respective feature vectors to generate respective plural candidate sentences in a second language; and selecting a final sentence in the second language from the respective plural candidate sentences in the second language.

Parallel processing-based translation method and apparatus

A speech signal recognition method, apparatus, and system. The speech signal recognition method may include obtaining by or from a terminal an output of a personalization layer, with respect to a speech signal provided by a user of the terminal, having been implemented by input of the speech signal to the personalization layer, the personalization layer being previously trained based on speech features of the user, implementing a global model by providing the obtained output of the personalization layer to the global model, the global model being configured to output a phonemic signal indicating a phoneme included in the speech signal through the global model being previously trained based on speech features common to a plurality of users, and re-training the personalization layer based on the phonemic signal output from the global model, where the personalization layer and the global model collectively represent an acoustic model.

Speech signal recognition system and method

A multilevel speech recognition method and an apparatus performing the method are disclosed. The method includes receiving a first speech command from a user through a speech interface, and extracting a keyword from the first speech command. The method also includes providing a candidate application group of a category providing a service associated with the keyword, and processing a second speech command from the user associated with an application selected from the candidate application group.

Multilevel speech recognition for candidate application group using first and second speech commands

A method and apparatus for speech recognition are provided. The method and the apparatus calculate signal to noise ratios (SNRs) of speech signals from a user received at speech recognition apparatuses. The method and the apparatus recognize a reference speech signal having a maximum SNR among the SNRs.

Growing community of inventors

Seoul, South Korea

Minyoung Mun