The processing of Japanese text is complicated by the fact that there are no word delimiters. To segment Japanese text, systems typically use knowledge-based methods and large lex...
We describe a new tagging model where the states of a hidden Markov model (HMM) estimated by unsupervised learning are incorporated as the features in a maximum entropy model. Our...
Automatic extraction of content description from commercial audio recordings has a number of important applications, from indexing and retrieval through to novel musicological ana...
To circumvent spam filters, many spammers attempt to obfuscate their emails by deliberately misspelling words or introducing other errors into the text. For example viagra may be...
We examine the use of hidden Markov and hidden semi-Markov models for automatically segmenting an electrocardiogram waveform into its constituent waveform features. An undecimated...
Nicholas P. Hughes, Lionel Tarassenko, Stephen J. ...