Exploring Unsupervised Features in Conditional Random Fields for Spanish Named Entity Recognition

Unsupervised features such as word representations mostly given by word embeddings have been shown significantly improve semi supervised Named Entity Recognition (NER) for English language. In this work we investigate whether unsupervised features can boost (semi) supervised NER in Spanish. To do so, we use word representations and collocations as additional features in a linear chain Conditional Random Field (CRF) classifier. Experimental results (82.44% F-score on the CoNLL-2002 corpus and 65.72% F-score on Ancora Corpus) show that our approach is comparable to some state-of-art Deep Learning approaches for Spanish, in particular when using cross-lingual Word Representations.