Corpora and concordancers on the nl.ijs.si server

The paper presents the monolingual and parallel corpora which can be accessed through two concordancers on the server nl.ijs.si. Twelve monolingual corpora contain Slovene language texts, one contains Japanese and one English texts, and comprise reference corpora, such as Gigafida for written contemporary Slovene, IMP for historical Slovene, and GOS for spoken Slovene and specialised corpora, such as the corpus of texts from the informatics domain and the corpus of Slovene tweets. The five parallel corpora contain Slovene texts sentence aligned with, variously, English, Japanese, French, German, and Italian from domains such as EU law, literature and journalism. Although most of the corpora have been produced in the past, they have now been newly annotated, some have been extended with additional texts, and a few are completely new. The texts in the corpora are supplied with meta-data, while their word tokens are either manually or automatically annotated with at least lemmas and morphosyntactic descriptions. Most of the corpora are freely available through two web concordancers, the noSketch Engine and CUWI. These two corpus analysis tools support searching large annotated corpora, various types of search result display, the possibility to filter the searches according to meta-data, and saving the search results locally. In addition to the corpora and concordancers the paper also discusses some issues pertaining to such a corpus-linguistic infrastructure, and concludes with directions for further work.