Confidence measures for large vocabulary continuous speech recognition

By Frank Wessel, Ralf Schlüter, Klaus Macherey, Hermann Ney. IEEE Transactions on Speech and Audio Processing, 2001

515 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/d3a51176b45a6b40b15d096a5cee605cddbe5644

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Confidence Estimation for Machine Translation (2004, background): “Many approaches, eg (Wessel et al., 2001), derive confidence scores, such as posterior probabilities P (y|x), directly from quantities in the base system.”
Lattice rescoring strategies for long short term memory language models in speech recognition (2017, methodology): “Both weights (Eqns 10 and 11) can be computed efficiently using a forward-backward algorithm [24].”
Parameter Estimation of Statistical Models Using Convex Optimization (2010, methodology): “Refer to [33] for details on running the forward-backward algorithm on a word graph.”
Enhanced Phone Posteriors for Improving Speech Recognition Systems (2010, background): “These long contextual and phonetic/lexical sources of knowledge can help in providing better phone posterior estimates; however, they are not usually taken into account in the MLP-based phone posterior estimation.”
Confidence measures for speech recognition: A survey (2005, methodology): “There are many other implementation issues to consider when computing posterior probability in word graph, e.g., scaling of probabilities in summation, elimination of redundant silence edges, etc. Readers are referred to Wessel et al. (2001) for more details.”
Learning Lexicons From Speech Using a Pronunciation Mixture Model (2013, background): “Another component found in some recognizers that can be efficiently represented using FSTs is a set of phonological rules [8].”
Large scale deep neural network acoustic modeling with semi-supervised training data for YouTube video transcription (2013, methodology): “, lattice-based word posteriors [13] could certainly be used instead of the simple binary match/no-match word-level measure used here.”
Unsupervised training of acoustic models for large vocabulary continuous speech recognition (2005, methodology): “For comparison, the Broadcast News’96 training corpus was also augmented with the Broadcast News’97 training corpus and the original, manually generated transcriptions.”