Abstract: A method for searching content having same voice as a voice of a target speaker from among a plurality of contents includes extracting a feature vector corresponding to the voice of the target speaker, selecting any subset of speakers from a training dataset repeatedly by a predetermined number of times, generating linear discriminant analysis (LDA) transformation matrices using each of the selected any subsets of speakers repeatedly by a predetermined number of times, projecting the extracted speaker feature vector to the selected corresponding subsets of speakers using each of the generated LDA transformation matrices, assigning a value corresponding to nearby speaker class among corresponding subsets of speakers, to each of projection regions of the extracted speaker feature vector, generating a hash value corresponding to the extracted feature vector based on the assigned values, and searching content having a similar hash value to the generated hash value among the contents.

Abstract: The present disclosure describes a system, method, and computer program for predicting sentiment labels for audio speech utterances using an audio speech sentiment classifier pretrained with pseudo sentiment labels. A speech sentiment classifier for audio speech (“a speech sentiment classifier”) is pretrained in an unsupervised manner by leveraging a pseudo labeler previously trained to predict sentiments for text. Specifically, a text-trained pseudo labeler is used to autogenerate pseudo sentiment labels for the audio speech utterances using transcriptions of the utterances, and the speech sentiment classifier is trained to predict the pseudo sentiment labels given corresponding embeddings of the audio speech utterances. The speech sentiment classifier is then subsequently fine tuned using a sentiment-annotated dataset of audio speech utterances, which may be significantly smaller than the unannotated dataset used in the unsupervised pretraining phase.

Abstract: A method for searching content having same voice as a voice of a target speaker from among a plurality of contents includes extracting a feature vector corresponding to the voice of the target speaker, selecting any subset of speakers from a training dataset repeatedly by a predetermined number of times, generating linear discriminant analysis (LDA) transformation matrices using each of the selected any subsets of speakers repeatedly by a predetermined number of times, projecting the extracted speaker feature vector to the selected corresponding subsets of speakers using each of the generated LDA transformation matrices, assigning a value corresponding to nearby speaker class among corresponding subsets of speakers, to each of projection regions of the extracted speaker feature vector, generating a hash value corresponding to the extracted feature vector based on the assigned values, and searching content having a similar hash value to the generated hash value among the contents.

Abstract: A method of converting a feature vector includes extracting a feature sequence from an audio signal including utterance of a user; extracting a feature vector from the feature sequence; acquiring a conversion matrix for reducing a dimension of the feature vector, based on a probability value acquired based on different covariance values; and converting the feature vector by using the conversion matrix.

Abstract: A method of converting a feature vector includes extracting a feature sequence from an audio signal including utterance of a user; extracting a feature vector from the feature sequence; acquiring a conversion matrix for reducing a dimension of the feature vector, based on a probability value acquired based on different covariance values; and converting the feature vector by using the conversion matrix.