MoRTy: Unsupervised Learning of Task-specialized Word Embeddings by Autoencoding

Nils Rethmeier, Barbara Plank

Publikation: Konference artikel i Proceeding eller bog/rapport kapitelKonferencebidrag i proceedingsForskningpeer review


Word embeddings have undoubtedly revolutionized NLP. However, pre-trained embeddings do not always work for a specific
task (or set of tasks), particularly in limited resource setups. We introduce a simple
yet effective, self-supervised post-processing
method that constructs task-specialized word
representations by picking from a menu of
reconstructing transformations to yield improved end-task performance (MORTY). The
method is complementary to recent state-of-the-art approaches to inductive transfer via
fine-tuning, and forgoes costly model architectures and annotation. We evaluate MORTY
on a broad range of setups, including different
word embedding methods, corpus sizes and
end-task semantics. Finally, we provide a surprisingly simple recipe to obtain specialized
embeddings that better fit end-tasks
TitelProceedings of the 4th Workshop on Representation Learning for NLP (RepL4NLP-2019)) : RepL4NLP-2019
ForlagAssociation for Computational Linguistics
ISBN (Elektronisk) 978-1-950737-35-2
StatusUdgivet - 2019


Dyk ned i forskningsemnerne om 'MoRTy: Unsupervised Learning of Task-specialized Word Embeddings by Autoencoding'. Sammen danner de et unikt fingeraftryk.