On letter to sound conversion for Romanian: A comparison of five algorithms

Author

Toma, Stefan-Adrian ; Birsan, Traian ; Totir, Felix ; Oancea, Eugeniu

Author_Institution

Mil. Tech. Acad., Bucharest, Romania

fYear

2013

fDate

16-19 Oct. 2013

Firstpage

1

Lastpage

6

Abstract

This paper presents an evaluation of 5 letter-to-sound (LTS) systems for Romanian. The first is an expert system; three of them use automatic classification methods with decision trees, neural networks and support vector machines respectively and the fifth one uses pronunciation by analogy. All systems were trained and tested on the same database: a 15,517 words corpus built according to the SpeechDat specifications and a corpus consisting of the most frequent 4779 words in Romanian. The results show that decision trees and neural networks generate the best results for letter to sound conversion in Romanian.

Keywords

decision trees; expert systems; neural nets; pattern classification; signal classification; speech processing; support vector machines; LTS systems; Romania; SpeechDat specifications; automatic classification methods; decision trees; expert system; letter-to-sound conversion systems; neural networks; pronunciation; support vector machines; word corpus; Context; Decision trees; Dictionaries; Error analysis; Support vector machine classification; Training; automatic classification; corpus; insert (key words); phonetic transcription; styling;

fLanguage

English

Publisher

ieee

Conference_Titel

Speech Technology and Human - Computer Dialogue (SpeD), 2013 7th Conference on

Conference_Location

Cluj-Napoca

Type

conf

DOI

10.1109/SpeD.2013.6682664

Filename

6682664