This web site is no longer maintained and the content may be outdated.
Please visit for up-to-date information.
No upcoming events...

Home / Graduate / PhD Theses Completed
  Haşim Sak, 2011  [download thesis]    

Thesis Title

Integrating Morphology into Automatic Speech Recognition: Morpholexical and Discriminative Language Models for Turkish


Languages with agglutinative or inflectional morphology prove to be challenging for speech and language processing due to relatively large vocabulary size leading to a high number of out-of-vocabulary (OOV) words. In this thesis, we tackle with these challenges in automatic speech recognition (ASR) frame for Turkish having an extremely productive inflectional and derivational morphology. First, we build the necessary tools and resources for Turkish, namely a nite-state morphological parser, a perceptron-based morphological disambiguator, and a text corpus collected from web. Second, we introduce two complementary language modeling approaches to alleviate OOV word problem and to exploit morphology as a knowledge source. The rst model, morpholexical language model, is a generative n-gram model, where modeling units are lexical-grammatical morphemes instead of commonly used words or statistical sub-words. We also propose a novel approach for integrating the morphology into an ASR system in the nite-state transducer framework as a knowledge source. The second model is a linear reranking model trained discriminatively with a variant of the perceptron algorithm, word error rate (WER) sensitive perceptron, using morpholexical and morphosyntactic features to rerank n-best candidates obtained with the generative model. We apply the proposed models in Turkish broadcast news transcription task and give experimental results. The morpholexical model is highly effective in alleviating OOV problem and improves the WER over word and statistical sub-word models by 1.8% and 0.8% absolute respectively. The discriminatively trained model further improves the WER of the system by 0.8% absolute. Finally, we present an algorithm for on-the-fly lattice rescoring with low-latency.
Boğaziçi University Department of Computer Engineering
Address: 34342 Bebek, Istanbul, TURKEY
Phone: +90 212 359 4523-24 Fax: +90 212 287 2461
general information:   webmaster: