A. Ljolje and F. Pereira and M. Riley. Efficient general lattice generation and rescoring. In Proceedings of the European Conference on Speech Communication and Technology (Eurospeech '99), Budapest, Hungary, 1999. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
Detecting Acoustic Morphemes in Lattices for.. -.. (Correct)
....best path 42 pruned 12 full lattice 6 Table 1: Comparison of the percentage of test sentences with no detected acoustic morphemes, in the best path, in the pruned lattice and in the full lattice experiments. Lattices are efficient representations of a distribution of alternative hypothesis [16] [17] To fix our ideas, a simple example of a lattice network, resulting from the utterance collect call , is shown in Figure 2, where a bold circle represents an initial state and a double circle a final state. The most likely phone sequence of the transcribed utterance is K ax l eh K T K ao ....
Llolje A., Pereira F. and Riley M.: Efficient General Lattice Generation and Rescoring, Proc. Eurospeech, Budapest, Hungary, vol 3, pp. 1251--1254, Sept. 1999.
Weighted Finite-State Transducers in Speech Recognition - Mohri, Pereira, Riley (2000) (28 citations) Self-citation (Pereira Riley) (Correct)
No context found.
A. Ljolje and F. Pereira and M. Riley. Efficient general lattice generation and rescoring. In Proceedings of the European Conference on Speech Communication and Technology (Eurospeech '99), Budapest, Hungary, 1999.
An Efficient Algorithm for the n-Best-Strings Problem - Mohri, Riley (2002) (1 citation) Self-citation (Riley) (Correct)
....utterance a lattice that represents the recognizer hypotheses as an acyclic weighted automaton. We then apply the n best algorithm in a separate step to this automaton. Since lattices with low lattice word error rates can be generated with very little added computation over one best recognition [8], this approach allows us to completely decouple the nbest generation from the first pass recognition step with negligible cost in accuracy or efficiency. We describe our algorithm in detail and report experimental results demonstrating its efficiency. We show that our algorithm is practical even ....
....hypothesis that best matches truth among the alternatives returned. The numbers along each curve in the figure correspond to beam widths 9 through 14 in steps of 1. By comparing the same beam widths on curve 1 and curve 5 in Figure 3, we see that this simple NAB system, very similar to the one in [8], produces lattices of good quality in time only a few percent slower than the 1 best result. By comparing the same beam widths on curve 5 and curve 4, we see that the added time for 1000 best generation from the lattice is negligible. In fact, the 1000 best generation experiment, which includes ....
A. Ljolje and F. Pereira and M. Riley, "Efficient general lattice generation and rescoring," in Proceedings of the European Conference on Speech Communication and Technology (Eurospeech '99), Budapest, Hungary, 1999.
The AT&T LVCSR-2000 System - Ljolje, Hindle, Riley, Sproat (2000) Self-citation (Ljolje Riley) (Correct)
....model 2. culling frequently seen alternative pronunciations in these transcriptions and adding them to the baseline lexicon along with their relative frequencies 6. SYSTEM ARCHITECTURE The recognition system used in the LVCSR 2000 evaluation is based on a cascade of rescoring of word lattices [5] using progressively more accurate models. There are many distinct steps in the recognition process. We will describe the overall strategy first. Later we will define each of the individual steps to complete the description of the system architecture. The overall system structure can be separated ....
A. Ljolje, F. Pereira, and M. Riley. Efficient General lattice Generation and Rescoring. In Proc. Eurospeech '99, Budapest, Hungary, 1999.
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