Nestandardno zapisivanje srpskog jezika na Tviteru: mnogo buke oko malo odstupanja?

This paper deals with non-standard spelling variants of Serbian words on Twitter. The analysis is based on a sample of automatically collected tweets identified as containing non-standard features. The sample was manually normalised, tagged with morhosyntactic descriptions and lemmatised. In the analysis, we compare non-standard forms to the standard ones they are related to and we establish the types of transformations that led to their formation. The results show that transformations are particularly frequent in closed-class words, such as auxiliary verbs, interjections and abbreviations. Character deletions are more common than character insertions and replacements. More transformations tend to occur at word end than in other positions. Despite the indisputable presence of non-standard forms on Serbian Twitter, we conclude that these forms are overall too infrequent and too functionally specialised to justify any claims about a negative influence of computer-mediated communication on the standard language norm.