Learn More
This paper presents a two level lexical stress assignment model for out of vocabulary Slovenian words used in our text-to-speech system. First, each vowel (and consonant 'r') is determined, whether it is stressed or un-stressed, and a type of lexical stress is assigned for every stressed vowel (and consonant 'r'). We applied a machine-learning technique(More)
We present an algorithm for parsing with detection of intra-clausal coordinations. The algorithm is based on machine learning techniques and helps to decompose a large parsing problem into several smaller ones. Its performance was tested on Slovene Dependency Treebank. Used together with the maximum spanning tree parsing algorithm it improved parsing(More)
The impact of clause and intraclausal coordination detection to dependency parsing of Slovene is examined. New methods based on machine learning and heuristic rules are proposed for clause and in-traclausal coordination detection. They were included in a new dependency parsing algorithm, PACID. For evaluation, Slovene dependency treebank was used. At(More)
The new Slovenian text-to-speech engine is a modular system consisting of four independent modules (text normalization, grapheme-to-phoneme conversion, prosody generation and segmental concatenation), which are pipelined together. Each module is responsible for one portion of the problem of converting from text into speech. That enables easy improvements of(More)
Accentuation of words is a basic task in the construction of any speech synthesis system. This task is particularly difficult in languages where lexical stress can be located almost arbitrarily on every syllable in the word, such as in the Slovene language. Therefore, we use machine learning methods to create accentuation rules. In this paper we compare the(More)
  • T. Sef
  • 2004
One of the characteristics of the Slovenian language is that lexical stress can be located almost arbitrarily on every syllable in the word, which makes the pronunciation very difficult. Some pronunciation rules exist, but their precision is not sufficient for efficient speech synthesis. Therefore a machine-learning technique (decision trees or boosted(More)