Korpusne oznake za opis konteksta govornih dogodkov v slovenskih govornih korpusih

The time-consuming and costly preparation of a speech corpus requires careful consideration of its composition and the categorization of the recorded metadata at the time of its design. The variety of speech events included in the national reference corpus should reflect the diversity of contemporary spoken language as much as possible. We will be interested in how to categorize the annotations used to describe the context of the speech events in order to achieve this representativeness without completely giving up the intercomparability of the data. A thoughtful design allows us to minimize the time-consuming tag adjustments required in subsequent corpus upgrades. We will carry out a comparative analysis of domestic and foreign speech corpora to critically evaluate the four basic categories of annotations used to describe the context of a speech situation. We will review the design of the foreign reference speech corpora FOLK, BNC2014, ORAL2013, the Spoken Dutch Corpus and C-ORAL-ROM and compare them with the current reference corpus of spoken Slovene Gos 2.1. We will problematize the selected annotations and highlight the more problematic areas that would require further consideration and potential future re-categorization.