We explore what effects word order has on cross-lingual sentiment classification. Specifically, we compare models that should be strongly influenced by word order (LSTMs), mildly affected (CNNs), and not affected (SVMs with Bag-of-embedding representations).
- Python 3
- NumPy
- sklearn [http://scikit-learn.org/stable/]
- keras [https://keras.io/]
- nltk [http://www.nltk.org/]
Copyright (C) 2018, Jeremy Barnes
Licensed under the terms of the Creative Commons CC-BY public license