Acoustic Modelling Using Continuous Rational Kernels Export

@article{citeulike:1614233, abstract = {Abstract\ \ Many discriminative classification algorithms are designed for tasks where samples can be represented by fixed-length vectors. However, many examples in the fields of text processing, computational biology and speech recognition are best represented as variable-length sequences of vectors. Although several dynamic kernels have been proposed for mapping sequences of discrete observations into fixed-dimensional feature-spaces, few kernels exist for sequences of continuous observations. This paper introduces continuous rational kernels, an extension of standard rational kernels, as a general framework for classifying sequences of continuous observations. In addition to allowing new task-dependent kernels to be defined, continuous rational kernels allow existing continuous dynamic kernels, such as Fisher and generative kernels, to be calculated using standard weighted finite-state transducer algorithms. Preliminary results on both a large vocabulary continuous speech recognition (LVCSR) task and the TIMIT database are presented.}, author = {Layton, Martin and Gales, Mark}, citeulike-article-id = {1614233}, citeulike-linkout-0 = {http://dx.doi.org/10.1007/s11265-006-0027-4}, citeulike-linkout-1 = {http://www.ingentaconnect.com/content/klu/vlsi/2007/00000048/F0020001/00000027}, day = {1}, doi = {10.1007/s11265-006-0027-4}, issn = {0922-5773}, journal = {The Journal of VLSI Signal Processing}, keywords = {acoustic\_models, asr, kernel\_methods, machine\_learning, rational\_kernels, sequence\_kernels, speech\_recognition}, month = {August}, number = {1}, pages = {67--82}, posted-at = {2009-09-01 00:02:02}, priority = {2}, publisher = {Springer}, title = {Acoustic Modelling Using Continuous Rational Kernels}, url = {http://dx.doi.org/10.1007/s11265-006-0027-4}, volume = {48}, year = {2007} }

TY - JOUR ID - citeulike:1614233 L3 - citeulike-article-id:1614233 N2 - Abstract  Many discriminative classification algorithms are designed for tasks where samples can be represented by fixed-length vectors. However, many examples in the fields of text processing, computational biology and speech recognition are best represented as variable-length sequences of vectors. Although several dynamic kernels have been proposed for mapping sequences of discrete observations into fixed-dimensional feature-spaces, few kernels exist for sequences of continuous observations. This paper introduces continuous rational kernels, an extension of standard rational kernels, as a general framework for classifying sequences of continuous observations. In addition to allowing new task-dependent kernels to be defined, continuous rational kernels allow existing continuous dynamic kernels, such as Fisher and generative kernels, to be calculated using standard weighted finite-state transducer algorithms. Preliminary results on both a large vocabulary continuous speech recognition (LVCSR) task and the TIMIT database are presented. IS - 1 JF - The Journal of VLSI Signal Processing SN - 0922-5773 EP - 82 TI - Acoustic Modelling Using Continuous Rational Kernels PB - Springer VL - 48 SP - 67 KW - acoustic_models KW - asr KW - kernel_methods KW - machine_learning KW - rational_kernels KW - sequence_kernels KW - speech_recognition AU - Layton, Martin AU - Gales, Mark PY - 2007/08/01/ UR - http://dx.doi.org/10.1007/s11265-006-0027-4 ER -