Regularized Word Embeddings in Text Classification

Use Python 3.4+ with Pip to install the required Python packages:

pip install -r requirements.txt

Reproducing Our Results

To reproduce our results, you can download all the datasets and corpora, produce Word2Vec models and similarity matrices, and perform the evaluation. Alternatively, you can download and visualize our result files.

Performing Your Own Evaluation

To perform your own evaluation, you will require the following additional tools: GNU Make, Perl 5, GNU Parallel, GNU Wget, Unzip, XZ Utils, GNU Coreutils, and Moreutils. Execute the following command:

dvc repro results.dvc

Open the Jupyter notebook with the experimental code to see the results:

jupyter-notebook classification.ipynb

Downloading and Visualizing Our Results

To download our results, execute the following command:

dvc pull results.dvc

Open the Jupyter notebook with the experimental code to see the results:

jupyter-notebook classification.ipynb

Citing

Text

NOVOTNÝ, Vít, Eniafe Festus AYETIRAN, Michal ŠTEFÁNIK and Petr SOJKA. Text classification with word embedding regularization and soft similarity measure. New York, USA: Cornell University, 2020.

BibTeX

@misc{novotny2020text,
  title = {{Text classification with word embedding regularization and soft similarity measure}},
  author = {V\'{i}t Novotn\'{y} and Eniafe Festus Ayetiran and Michal \v{S}tef\'{a}nik and Petr Sojka},
  year = 2020,
  eprint = {2003.05019v1},
  archivePrefix = {arXiv},
  primaryClass = {cs.IR},
  url = {https://arxiv.org/abs/2003.05019v1},
}