FST-Based Pronunciation Lexicon Compression for Speech Engines

Abstract

Finite‐state transducers are frequently used for pronunciation lexicon representations in speech engines, in which memory and processing resources are scarce. This paper proposes two possibilities for further reducing the memory footprint of finite‐state transducers representing pronunciation lexicons. First, different alignments of grapheme and allophone transcriptions are studied and a reduction in the number of states of up to 30% is reported. Second, a combination of grapheme‐to‐allophone rules with a finite‐state transducer is proposed, which yields a 65% smaller finite‐state transducer than conventional approaches

Mlphon: A Multifunctional Grapheme-Phoneme Conversion Tool Using Finite State Transducers

“…When it enters the final state, the FST would have accepted every symbol in the input string. The output string is made up of all the emitted symbols at that point [50]. For the cases where the number of symbols in input or output strings mismatch, a 'null symbol', ɛ is introduced in the transition mapping.…”

Zgoščena predstavitev slovarjev izgovarjav s končnimi super pretvorniki

“…V literaturi je mogoče zaslediti tri osnovne pristope, ki omogočajo zgoščeno predstavitev slovarjev izgovarjav, in sicer z uporabo oštevilčenih končnih avtomatov (Lucchesi in Kowaltowski, 1993;Daciuk in Piskorski, 2011), dreves predpon (Ristov, 2005) ter z uporabo končnih pretvornikov (Mohri, 1994;Golob at al., 2012). V članku bomo predstavili nov način zgoščene predstavitve slovarjev izgovarjav z uporabo t. i. končnih super pretvornikov, ki predstavljajo nadgradnjo običajnih končnih pretvornikov.…”