Гибридное представление слова/части речи

We present a method that improves data selection by combining a hybrid word/part-of-speech representation for corpora, with the idea of distinguishing between rare and frequent events. We validate our approach …...