Лексические встраивания

AbstractLexical embeddings can serve as useful representations for words for a variety of NLP tasks, but learning embeddings for phrases can be challenging. While separate embeddings are learned for each …...

AbstractWe consider the extent to which different deep neural network (DNN) configurations can learn syntactic relations, by taking up Linzen et al.’s (2016) work on subject-verb agreement with LSTM RNNs. …...

Large generative language models have been very successful for English, but other lan- guages lag behind, in part due to data and com- putational limitations. We propose a method that …...