Provides a capsnet implementation for text classification
We provide the following notebooks:
- dataset-to-embeddings.ipynb: to transform headlines to embedding vectors used as new dataset to perform classification
- dataset-analysis.ipynb: to obtain general information regarding the dataset: most frequent terms, which cannot be represented as embeddings, most frequent words particular to a specific topic, etc.
- XGBoost.ipynb: classification performed with XGBoost
- ConvRec.ipynb: ConvRec implementation and dataset classification
- CapsNet.ipynb: CapsNet implementation and dataset classification
Results we obtained:
| Metric | CapsNet | XGBoost | ConvRec |
|---|---|---|---|
| Accuracy | 0.8901 | 0.8179 | 0.8897 |
| Time trained | 920 minutes | 23 minutes | 189 minutes |
To start a Docker image, run:
- docker run -p 8888:8888 -v "$PWD":/home/jovyan jupyter/datascience-notebook
- docker exec b7f3abbf54da pip install keras