Adapting word2vec to Named Entity Recognition
Scharolta Katharina Sienċnik · 2015
In this paper we explore how word vectors built using word2vec can be used to improve the performance of a classifier during Named Entity Recognition. Thereby, we discuss the best integration of word embeddings into the classification problem and consider the effect of the size of the unlabelled dataset on performance, reaching the unexpected result that for this particular task increasing the amount of unlabelled data does not necessarily increase the performance of the classifier.