Citation Context

...ting the named entity classification, there were used both a subset of the labeled version of the Bosque (≈ 20k tokens) and the Corpus-Web (with about 55k tokens of different varieties of Portuguese) =-=[9]-=-. In order to build the gazetteers, the Portuguese version of the Wikipedia was used for extracting entity names. Apart from that, large lists of countries and cities were also merged, together with t...