We used deep-learning methods to perform multi-label text classification on the Reuters-21578 Text Categorization Collection.
The architectures explored and their performances are shown below:
| Architecture | F1 Score |
|---|---|
| CNN | 0.7655879 |
| Dense | 0.6601627 |
| RNN (LSTM+Attention) | 0.1065762 |
A full report of our findings can be found at https://paul-tqh-nguyen.github.io/reuters_topic_labelling/.
The tools utilized in our exploration include:
If you're interested in more of my work, see my website.
Thanks for visiting!