Learning Label Embeddings for Nearest-Neighbor Multi-class Classification with an Application to Speech Recognition
نویسندگان
چکیده
We consider the problem of using nearest neighbor methods to provide a conditional probability estimate, P (y|a), when the number of labels y is large and the labels share some underlying structure. We propose a method for learning label embeddings (similar to error-correcting output codes (ECOCs)) to model the similarity between labels within a nearest neighbor framework. The learned ECOCs and nearest neighbor information are used to provide conditional probability estimates. We apply these estimates to the problem of acoustic modeling for speech recognition. We demonstrate significant improvements in terms of word error rate (WER) on a lecture recognition task over a state-of-the-art baseline GMM model.
منابع مشابه
ML-KNN: A lazy learning approach to multi-label learning
Multi-label learning originated from the investigation of text categorization problem, where each document may belong to several predefined topics simultaneously. In multi-label learning, the training set is composed of instances each associated with a set of labels, and the task is to predict the label sets of unseen instances through analyzing training instances with known label sets. In this...
متن کاملA Comparative Study of Gender and Age Classification in Speech Signals
Accurate gender classification is useful in speech and speaker recognition as well as speech emotion classification, because a better performance has been reported when separate acoustic models are employed for males and females. Gender classification is also apparent in face recognition, video summarization, human-robot interaction, etc. Although gender classification is rather mature in a...
متن کاملSupervised Manifold Learning with Incremental Stochastic Embeddings
In this paper, we introduce an incremental dimensionality reduction approach for labeled data. The algorithm incrementally samples in latent space and chooses a solution that minimizes the nearest neighbor classification error taking into account label information. We introduce and compare two optimization approaches to generate supervised embeddings, i.e., an incremental solution construction ...
متن کاملA Coupled k-Nearest Neighbor Algorithm for Multi-label Classification
ML-kNN is a well-known algorithm for multi-label classification. Although effective in some cases, ML-kNN has some defect due to the fact that it is a binary relevance classifier which only considers one label every time. In this paper, we present a new method for multi-label classification, which is based on lazy learning approaches to classify an unseen instance on the basis of its k nearest ...
متن کاملPresentation of K Nearest Neighbor Gaussian Interpolation and comparing it with Fuzzy Interpolation in Speech Recognition
Hidden Markov Model is a popular statisical method that is used in continious and discrete speech recognition. The probability density function of observation vectors in each state is estimated with discrete density or continious density modeling. The performance (in correct word recognition rate) of continious density is higher than discrete density HMM, but its computation complexity is very ...
متن کامل