Leveraging a Spam Text Classification example, different word embedding techniques were compared. The first 3 were paired with a simple Random Forest.
Precision | Recall | Accuracy | |
---|---|---|---|
TF-IDF | 100% | 83.1% | 97.8% |
word2vec | 60% | 22.3% | 87.7% |
doc2vec | 86.8% | 44.6% | 91.7% |
RNN | 98.7% | 96.9% | 93.3% |