Sequential subspace alignment of word embeddings
This repository contains experiments and visualizations for temporal domain adaptation in the form of semi-supervised subspace alignment for word embeddings. It accompanies the paper
"Back to the future -- Sequential alignment of text representations"
which is to be presented at the AAAI Conference on Artificial Intelligence, 2020 (preprint).
Natural Language Processing tasks currently tackled:
- annual paper acceptance prediction
- temporal named entity recognition
- rumour stance and veracity prediction.
Data
Rumour stance data stems from the EU project PHEME, which revolves around automatically determining the truth value of online statements. Specifically, we're looking at the journalism use case.
Contact
Questions and comments can be submitted to the issues tracker.