Title: The Ukrainian broadcast speech corpus. 
In this paper a short description of activities towards building 
a media-based speech corpus for spoken Ukrainian language is given. 
Different roles and specific features of text corpus and speech corpus 
are investigated as well as the most frequent mistakes and misunderstandings 
of the concept of a speech corpus are mentioned. 
The concept of a big representative corpus of spoken language and 
its desired properties are presented. 
The paper gives an overview of the current state of the art in speech corpora all over the world. 
It explains the need for a national speech corpus and indicates some 
of the typical areas of research and applications taking advantage of the existence of such a corpus. 
The speech databases currently available in Ukraine 
are listed and the particularities of annotation structures of these databases are pointed out. 
The authors search for a general annotation structure suitable for the kind of speech corpus envisaged. 
Some of the basic concepts and technical solutions used in recording and 
computer aided annotation used for the existing speech corpora are described. 
The most significant problems standing in the way of building a big speech corpus are pointed out. 
Furthermore, a pilot version of a speech corpus is presented, containing several recordings and 
their orthographic transcription. 
Index items: speech corpus, database, broadceast speech, spoken language, Ukrainian.