• transfer learning
• neural networks
• part-of-speech tagging
• deep learning
• multitask models
• natural language processing
• recursive neural networks
• morphological disambiguation

In this thesis, we address the task of morphological disambiguation in morphologically rich languages. In particular, we implement neural models using a limited amount of available labeled data. This is a central problem in many endangered languages. We apply deep learning techniques to solve morphological ambiguity relying on existing morphological analyzers. We consider the problem of disambiguating the part-of-speech and the lemma of ambiguous words, given the context of the words. The idea is to train recurrent neural networks to understand the context and to help us discriminate between the analyzer's options.
We evaluate single-task models and multi-task models and we achieve state-of-art accuracy for Italian, Russian, and Finnish morphological ambiguity.