G. Williams. A Study of the Use and Evaluation of Confidence Measures in Automatic Speech Recognition. Technical report, University of Sheffield, March 1998.  Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Rejection Measures for Handwriting Sentence Recognition - Marukatat, Artieres.. (2002) (4 citations) (Correct)
....model and to determine the most likely sequence of words (i.e. recognized sentence) 3. Confidence Measures There are many ways to define a confidence measure. Some methods that have been proposed exploit different features obtained from the decoding step, for example the lattice density [7, 16], N bests list [14] or language models [15] etc. Some other works use a post classifier to combine features such as likelihood and other statistics gathered from the decoding process (e.g. the number of letters in word, etc. into one measure [6, 8, 13] By far, however, the most popular ....
G. Williams. A Study of the Use and Evaluation of Confidence Measures in Automatic Speech Recognition. Technical report, University of Sheffield, March 1998.
Summarisation Of Spoken Audio Through Information Extraction - Valenza, Robinson, al. (1999) (4 citations) (Correct)
.... in calculating acoustic confidence measures [6] 11] Gethin Williams and Steve Renals at the University of Sheffield report using acoustic based confidence measures derived from the posterior phone probabilities of the Abbot HMM ANN speech recogniser to verify speech recognition hypotheses [9] [10]. They define the duration normalised posterior probability of a hypothesised word as the product of the posterior probability estimates of the constituent phones. Williams and Renals determined experimentally that they achieved the most discriminating measures by normalising all phones in a word ....
....probability estimates of the constituent phones. Williams and Renals determined experimentally that they achieved the most discriminating measures by normalising all phones in a word by the duration of the entire hypothesised word, rather than normalising each phone constituent by its own duration [10]: CM npost (q k ) 1 n e Gamma n s ne X n=ns log(p(q k jx n ) 2) CM npost = CM post (q k ) n e Gamma n s (3) This confidence measure has the additional advantage of being straightforward to compute with values available directly from the decoder output. For this study, we ....
Williams, G. " A Study of the Use and Evaluation of Confidence Measures in Automatic Speech Recognition. " Technical report, University of Sheffield University, Dept. of Computer Science, 1998.
Not just what, but also when: Guided automatic.. - Fosler-Lussier, Williams Self-citation (Williams) (Correct)
....used to rescore n best lists, facilitating the use of complex pronunciation models. One of the benefits of using a connectionist acoustic model in conjunction with our FSG decoder is the availability of posterior probability based acoustic confidence measures at both the phone and the word levels [8]. We found these acoustic confidence measures useful on several occasions as a guide to model building, including selection of pronunciations and checking automatically created models for novel words. Since the acoustic model was being developed at the same time as the pronunciation models, we ....
Gethin Williams. A study of the use and evaluation of confidence measures in automatic speech recognition. Technical Report CS-98-02, Department of Computer Science, University of Sheffield, 1998.
Confidence Measures for Evaluating Pronunciation Models - Williams, Renals (1998) (2 citations) Self-citation (Williams) (Correct)
....In this paper we make use of an acoustic confidence measure based on local estimates of posterior phone probabilities, CM npost . In previous studies, we have compared the performance of CM npost to that of a number of other confidence measures for the task of decoding hypothesis verification [12, 13]. We have found CM npost to perform better than the other confidence measures for the task of phone hypothesis verification and to be the least expensive to compute. A description of CM npost and four other confidence measures is given below for a phone q k with an hypothesised start time n s ....
G. Williams. A Study the Use and Evaluation of Confidence Measures in Automatic Speech Recognition. Technical Report, CS-98-02, Dept. Comp. Sci., Sheffield University.
Confidence Measures Derived From An Acceptor HMM - Williams, Renals (1998) (1 citation) Self-citation (Williams) (Correct)
....task dependent results. In addition to unconditional and conditional error rates, a range of evaluation metrics including mutual information [4] ROC (Receiver Operating Characteristic) 11] and DET (Detection Error Tradeoff) 6] curves and distributional separability have been investigated in [10]. Two findings of this investigation were that the diverse set of metrics broadly agree in their evaluations and that duration normalisation was beneficial for all confidence measures. Utterance verification experiments were performed using the Hub3 1995 evaluation test set of the NAB corpus and ....
G. Williams "A study of the use and evaluation of confidence measures in automatic speech recognition ". Technical report CS-98-02, Department of Computer Science, University of Sheffield, 1998. http://www.dcs.shef.ac.uk/people/G.Williams.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC