@InProceedings{sak-et-al-gotal-08,
Abstract = {In this paper, we propose a set of language resources
for building Turkish language processing applications.
Specifically, we present a finite-state implementation of a morphological parser,
an averaged perceptron-based morphological disambiguator,
and compilation of a web corpus. Turkish is an agglutinative
language with a highly productive inflectional and derivational morphology.
We present an implementation of a morphological parser based on two-level morphology.
This parser is one of the most complete parsers for Turkish
and it runs independent of any other external system
such as PC-KIMMO in contrast to existing parsers.
Due to complex phonology and morphology of Turkish, parsing introduces some ambiguous parses.
We developed a morphological disambiguator with accuracy of about 98%
using averaged perceptron algorithm. We also present our efforts
to build a Turkish web corpus of about 423 million words.},
Author = {Ha{\c s}im Sak and Tunga G{\"u}ng{\"o}r and Murat Sara{\c c}lar},
Booktitle = {GoTAL 2008},
Pages = {417--427},
Title = {Turkish Language Resources: Morphological Parser, Morphological Disambiguator and Web Corpus},
Volume = {5221},
Series = {LNCS},
Publisher = {Springer},
Year = {2008}
}
@inproceedings{arisoy-et-al-interspeech-07,
Abstract = {The aim of this study is to develop a speech recognition system
for Turkish broadcast news. State-of-the-art speech recognition
systems utilize statistical models. A large amount of data is required
to reliably estimate these models. For this study, a large
Turkish Broadcast News database, consisting of the speech signal
and corresponding transcriptions, is being collected. In this
paper, information about this database and experiments performed
using the system developed on the collected data are
presented. In addition to the baseline system, various sub-word
language models are investigated. Lexical stem-endings are
proposed as a novel unit for language modeling and are shown
to perform better than surface stem-endings and morphs. Currently,
our best systems have lower than 20% error on clean speech.},
Author = {Ebru Ar{\i }soy and Ha{\c s}im Sak and Murat Sara{\c c}lar},
Booktitle = {Proceedings of Interspeech 2007 - Eurospeech (To appear)},
Title = {Language Modeling for Automatic {Turkish} Broadcast News Transcription},
Year = {2007},
}
@inproceedings{sak-et-al-cicling-07,
Abstract = {This paper describes the application of the perceptron algorithm
to the morphological disambiguation of Turkish text. Turkish has
a productive derivational morphology. Due to the ambiguity caused by
complex morphology, a word may have multiple morphological parses,
each with a different stem or sequence of morphemes. The methodology
employed is based on ranking with perceptron algorithm which has
been successful in some NLP tasks in English. We use a baseline statistical
trigram-based model of a previous work to enumerate an n-best list
of candidate morphological parse sequences for each sentence. We then
apply the perceptron algorithm to rerank the n-best list using a set of
23 features. The perceptron trained to do morphological disambiguation
improves the accuracy of the baseline model from 93.61% to 96.80%.
When we train the perceptron as a POS tagger, the accuracy is 98.27%.
Turkish morphological disambiguation and POS tagging results that we
obtained is the best reported so far.},
Author = {Ha{\c s}im Sak and Tunga G{\"u}ng{\"o}r and Murat Sara{\c c}lar},
Booktitle = {CICLing 2007},
Pages = {107--118},
Title = {Morphological Disambiguation of {Turkish} Text with Perceptron Algorithm},
Volume = {LNCS 4394},
Year = {2007},
Url = {http://www.cmpe.boun.edu.tr/~hasim/papers/CICLing07.pdf}}
@article{sak-et-al-turkjelec-06,
Abstract = {Speech synthesis is the process of converting written text into machine-generated synthetic speech.
Concatenative speech synthesis systems form utterances by concatenating pre-recorded speech units.
Corpus-based methods use a large inventory to select the units to be concatenated. In this paper, we
design and develop an intel ligible and natural sounding corpus-based concatenative speech synthesis system
for the Turkish language. The implemented system contains a front-end comprised of text analysis,
phonetic analysis, and optional use of transplanted prosody. The unit selection algorithm is based on
commonly used Viterbi decoding algorithm of the best-path in the network of the speech units using spectral
discontinuity and prosodic mismatch objective cost measures. The back-end is the speech waveform
generation based on the harmonic coding of speech and overlap-and-add mechanism. Harmonic coding
enabled us to compress the unit inventory size by a factor of three. In this study, a Turkish phoneme
set has been designed and a pronunciation lexicon for root words has been constructed. The importance
of prosody in unit selection has been investigated by using transplanted prosody. A Turkish Diagnostic
Rhyme Test (DRT) word list that can be used to evaluate the intel ligibility of Turkish Text-to-Speech
(TTS) systems has been compiled. Several experiments have been performed to evaluate the quality of the
synthesized speech and we obtained 4.2 Mean Opinion Score (MOS) in the listening tests for our system,
which is the first unit selection based system published for Turkish.},
Author = {Ha{\c s}im Sak and Tunga G{\"u}ng{\"o}r and Ya{\c s}ar Safkan},
Date-Modified = {2007-07-29 19:01:48 +0300},
Journal = {Turkish Journal of Electrical Engineering and Computer Sciences},
Number = {2},
Pages = {209--223},
Title = {A Corpus-Based Concatenative Speech Synthesis System for {Turkish}},
Volume = {14},
Year = {2006},
Url = {http://www.cmpe.boun.edu.tr/~hasim/papers/TurkJElecEngin06.pdf}}
@inproceedings{sak-et-al-eusipco-05,
Abstract = {In this paper, we design and develop an intelligible and natural
sounding corpus-based concatenative speech synthesis system for
Turkish. The implemented system contains a front-end comprised
of text analysis, phonetic analysis, and optional use of transplanted
prosody. The unit selection algorithm is based on commonly used
Viterbi decoding algorithm. The back-end is the speech waveform
generation based on the harmonic coding of speech and overlap-and-add
mechanism. In this study, a Turkish phoneme set has been
designed and a pronunciation lexicon for root words has been
constructed. For assessing the intelligibility of the synthesized
speech, a DRT word list for Turkish has been compiled. The
developed system obtained 4.2 Mean Opinion Score (MOS) in the
listening tests.},
Author = {Ha{\c s}im Sak and Tunga G{\"u}ng{\"o}r and Ya{\c s}ar Safkan},
Booktitle = {13th European Signal Processing Conference (EUSIPCO 2005)},
Title = {Generation of Synthetic Speech from {Turkish} Text},
Year = {2005},
Url = {http://www.cmpe.boun.edu.tr/~hasim/papers/EUSIPCO05.pdf}}
@mastersthesis{sak-msthesis-04,
Abstract = {Speech synthesis (text-to-speech) is the process of converting the written text
into machine generated synthetic speech. Concatenative speech synthesis systems render
speech by concatenating pre-recorded speech units. Corpus-based methods (unit
selection) use a large inventory to select the units and concatenate. This thesis is part
of an effort to design and develop an intelligible and natural sounding corpus-based
concatenative speech synthesis system for Turkish. The implemented system contains
a relatively simple front-end comprised of text analysis, phonetic analysis, and optional
use of transplanted prosody. The unit selection algorithm is based on commonly used
Viterbi decoding algorithm of the best path in the network of the units. The back-end
is the speech waveform generation based on the harmonic coding of speech and overlapand-add mechanism.
In this work, the different unit sizes such as syllables, phones and
half-phones have been experimented with. Speech corpus design and recording script
preparation methods have been explained. A speech model based on harmonic coding
of speech has been developed for speech representation and waveform generation. The
harmonic coding has enabled us to compress the unit inventory size by a factor of three.
A Viterbi decoding algorithm using spectral discontinuity cost and prosodic mismatch
ob jective cost measures has been implemented. A Turkish phoneme set has been designed.
Text-to-phoneme conversion for Turkish has been worked on, and a root words
pronunciation lexicon has been constructed. A simple text normalization module has
been implemented. The importance of prosody in unit selection has been studied by
using transplanted prosody vs no synthetic prosody modeling in unit selection. Subjective
tests have been carried out for evaluating the synthesized speech quality. The
final Turkish speech synthesis system got 4.2 MOS like score in the listening tests.},
Author = {Ha{\c s}im Sak},
Date-Modified = {2007-07-29 19:20:05 +0300},
School = {Bo{\u g}azi{\c c}i University},
Title = {A Corpus-Based Concatenative Speech Synthesis System for {Turkish}},
Year = {2004},
Url = {http://www.cmpe.boun.edu.tr/~hasim/papers/MSThesis.pdf}}