Katarzyna Klessa

Polish unit selection speech synthesis with BOSS: extensions and speech corpora

This article presents research and development aimed at creating a Polish speech database for speech synthesis and adapting BOSS (The Bonn Open Synthesis System) to the Polish language. First of all, the linguistic background for the design of Polish spoken resources for unit selection is presented, together with the presentation of the applied transcription and annotation methods. The next section details the assumptions and the structure of the Polish corpus and its segmental and prosodic annotation. Then, the linguistic features used in duration modelling and the selection of adequate speech units of two Polish modules in BOSS are reported: the duration prediction module (the description is accompanied by a concise overview of Polish duration modelling for speech technology purposes) and the cost functions module. Finally, the results of two kinds of perception tests are discussed: the first is a preference test aimed at the evaluation of synthesized speech obtained using three variants of speech signal segmentation (automatic, semi-automatic and manual) and the second is a mean opinion score test carried out to provide a preliminary assessment of the synthesized speech quality attained with the Polish version of the BOSS synthesizer. The closing chapter summarizes future perspectives and challenges for the Polish TTS (text-to-speech) and further developments of BOSS for Polish.

Duration and speed of speech events: A selection of methods

Abstract The study of speech timing, i.e. the duration and speed or tempo of speech events, has increased in importance over the past twenty years, in particular in connection with increased demands for accuracy, intelligibility and naturalness in speech technology, with applications in language teaching and testing, and with the study of speech timing patterns in language typology. H owever, the methods used in such studies are very diverse, and so far there is no accessible overview of these methods. Since the field is too broad for us to provide an exhaustive account, we have made two choices: first, to provide a framework of paradigmatic (classificatory), syntagmatic (compositional) and functional (discourse-oriented) dimensions for duration analysis; and second, to provide worked examples of a selection of methods associated primarily with these three dimensions. Some of the methods which are covered are established state-of-the-art approaches (e.g. the paradigmatic Classification and Regression Trees, CART , analysis), others are discussed in a critical light (e.g. so-called ‘rhythm metrics’). A set of syntagmatic approaches applies to the tokenisation and tree parsing of duration hierarchies, based on speech annotations, and a functional approach describes duration distributions with sociolinguistic variables. Several of the methods are supported by a new web-based software tool for analysing annotated speech data, the Time Group Analyser.