Augmenting a BiLSTM tagger with a Morphological Lexicon and a Lexical Category Identification Step

Previous work on using BiLSTM models for PoS tagging has primarily focused on\nsmall tagsets. We evaluate BiLSTM models for tagging Icelandic, a\nmorphologically rich language, using a relatively large tagset. Our baseline\nBiLSTM model achieves higher accuracy than any previously published tagger not\ntaking advantage of a morphological lexicon. When we extend the model by\nincorporating such data, we outperform previous state-of-the-art results by a\nsignificant margin. We also report on work in progress that attempts to address\nthe problem of data sparsity inherent in morphologically detailed, fine-grained\ntagsets. We experiment with training a separate model on only the lexical\ncategory and using the coarse-grained output tag as an input for the main\nmodel. This method further increases the accuracy and reduces the tagging\nerrors by 21.3% compared to previous state-of-the-art results. Finally, we\ntrain and test our tagger on a new gold standard for Icelandic.\n