Some facts about myself
My name is Andreas and I am currently an R&D Data Scientist at Mudano.
I recently finished my Masters at the University of Edinburgh where I specialised in Natural Language Processing and applied Machine Learning.
More specifically, I learned a lot about the dark art of training neural networks and became familiar with the language modelling and dependency parsing literature.
My dissertation, supervised by Adam Lopez and Clara Vania, was on analysing the syntactic structure of sentences using neural network models for languages with rich morphology (demo, code). The results confirmed that for such languages, encoders which construct word representations from characters significantly outperform others that model the input on the word level.
I am currently working on understanding what patterns neural network character encoders capture in language and on exploring the effect that multi-task learning can have on the learned representations.
Before that I worked on named entity recognition for Greek and Serbian under the supervision of Iraklis Varlamis.
I graduated from Harokopio University of Athens in 2014. For my B.Sc. dissertation, supervised by Rania Hatzi, Mara Nikolaidou and Dimosthenis Anagnostopoulos, I worked on a word similarity measure with the intent of discerning whether two sentences could be aligned - were on the same topic. The word representations I constructed were influenced by the lectures - reading of the book Gödel, Escher, Bach.