hrvatski

Priprema usporedivih korpusa za usporedbu

Although the six corpora included in the research were comparable with respect to size, purpose and structure, it was indispensable, due to the nature of the planned quantitative analysis, to make them comparable at the POS and MSD tagging level. Since the tagsets used to annotate the corpora were only partially compatible, several procedures were needed to convert the existing tags to a common tagset in order to have comparable results. However, also in case of full compatibility with international standards, it is inevitable to think about and to compare the tagsets because of the differences in the perception and in the existence of grammatical categories in different languages, i.e. Croatian and Italian. After the differences among the tagsets of the six corpora were identified, followed by a detailed contrastive analysis of the two languages and after the only possible common POS and MSD tagset was found, the normalization of the corpora was performed. In order to achieve better comparability of results at inter-lingual level only the distribution within the common, comparable and relevant tags were taken into account which contributed to greater reliability and accuracy of results. On one hand this paper confirmed the importance of systematic planning of linguistic annotation scheme for each language in accordance with guidelines which prescribe international standards and create conditions for the comparability across corpora at both inter-lingual and intra-lingual levels. On the other hand, the paper showed that comparing and analysing MSD or POS tagsets can be considered a good basis and an interesting approach for the contrastive analysis.

usporedivi korpusi; kontrastivna analiza; POS označavanje; korpusna lingvistika

nije evidentirano