Optimization of Weighted Finite State Transducer for Speech Recognition

A scalable large vocabulary, speaker independent speech recognition system is being developed using Hidden Markov Models (HMMs) for acoustic modeling and a Weighted Finite State Transducer (WFST) to compile sentence, word, and phoneme models. The system comprises a software backend search and an FPGA-based Gaussian calculation which are covered here. In this paper, we present an efficient pipelined design implemented both as an embedded peripheral and as a scalable, parallel hardware accelerator. Both architectures have been implemented on an Alpha Data XRC-5T1, reconfigurable computer housing a Virtex 5 SX95T FPGA. The core has been tested and is capable of calculating a full set of Gaussian results from 3825 acoustic models in 9.03 ms which coupled with a backend search of 5000 words has provided an accuracy of over 80%. Parallel implementations have been designed with up to 32 cores and have been successfully implemented with a clock frequency of 133 MHz.

Download Full-text

BORDERS AND FINITE AUTOMATA

International Journal of Foundations of Computer Science ◽

10.1142/s0129054107005029 ◽

2007 ◽

Vol 18 (04) ◽

pp. 859-871

Author(s):

MARTIN ŠIMŮNEK ◽

BOŘIVOJ MELICHAR

Keyword(s):

Pattern Matching ◽

Hamming Distance ◽

Finite Automata ◽

Music Analysis ◽

Theoretical Description ◽

Specific Form ◽

Distance Measures ◽

Computer Assisted ◽

Finite State ◽

Finite State Transducer

A border of a string is a prefix of the string that is simultaneously its suffix. It is one of the basic stringology keystones used as a part of many algorithms in pattern matching, molecular biology, computer-assisted music analysis and others. The paper offers the automata-theoretical description of Iliopoulos's ALL_BORDERS algorithm. The algorithm finds all borders of a string with don't care symbols. We show that ALL_BORDERS algorithm is an implementation of a finite state transducer of specific form. We describe how such a transducer can be constructed and what should be the input string like. The described transducer finds a set of lengths of all borders. Last but not least, we define approximate borders and show how to find all approximate borders of a string when we concern Hamming distance definition. Our solution of this problem is based on transducers again. This allows us to use analogy with automata-based pattern matching methods. Finally we discuss conditions under which the same principle can be used for other distance measures.

Download Full-text

New speech recognition approaches based upon finite state vector quantization with structural constraints

10.1109/icassp.1988.196544 ◽

2003 ◽

Cited By ~ 1

Author(s):

P.-y. Ting ◽

C.-y. Tseng ◽

L.-s. Lee

Keyword(s):

Speech Recognition ◽

Vector Quantization ◽

State Vector ◽

Structural Constraints ◽

Finite State

Download Full-text

FINITELY SUBSEQUENTIAL TRANSDUCERS

International Journal of Foundations of Computer Science ◽

10.1142/s0129054103002126 ◽

2003 ◽

Vol 14 (06) ◽

pp. 983-994 ◽

Cited By ~ 9

Author(s):

CYRIL ALLAUZEN ◽

MEHRYAR MOHRI

Keyword(s):

Speech Recognition ◽

Finite Number ◽

Efficient Algorithm ◽

Experimental Results ◽

Theoretical Formulation ◽

Large Vocabulary ◽

Finite State Transducers ◽

Finite State ◽

Large Vocabulary Speech Recognition

Finitely subsequential transducers are efficient finite-state transducers with a finite number of final outputs and are used in a variety of applications. Not all transducers admit equivalent finitely subsequential transducers however. We briefly describe an existing generalized determinization algorithm for finitely subsequential transducers and give the first characterization of finitely subsequentiable transducers, transducers that admit equivalent finitely subsequential transducers. Our characterization shows the existence of an efficient algorithm for testing finite subsequentiability. We have fully implemented the generalized determinization algorithm and the algorithm for testing finite subsequentiability. We report experimental results showing that these algorithms are practical in large-vocabulary speech recognition applications. The theoretical formulation of our results is the equivalence of the following three properties for finite-state transducers: determinizability in the sense of the generalized algorithm, finite subsequentiability, and the twins property.

Download Full-text

Bootstrapping a Neural Morphological Generator from Morphological Analyzer Output for Inuktitut

10.33011/computel.v2i.455 ◽

2019 ◽

Vol 2 (1) ◽

Author(s):

Jeffrey Micher

Keyword(s):

Neural Network ◽

Training Data ◽

Data Set ◽

Set Size ◽

The Neural Network ◽

Surface Character ◽

Finite State ◽

Character Sequences ◽

Finite State Transducer

We present a method for building a morphological generator from the output of an existing analyzer for Inuktitut, in the absence of a two-way finite state transducer which would normally provide this functionality. We make use of a sequence to sequence neural network which “translates” underlying Inuktitut morpheme sequences into surface character sequences. The neural network uses only the previous and the following morphemes as context. We report a morpheme accuracy of approximately 86%. We are able to increase this accuracy slightly by passing deep morphemes directly to output for unknown morphemes. We do not see significant improvement when increasing training data set size, and postulate possible causes for this.

Download Full-text