Averaging Vectors + Random Forest
Word Embedding |
Details |
Score |
Bag of Words |
Tutorial. Remove stop words, numbers. |
0.84644 |
Word2Vec skip-gram |
Tutorial. Spell out digits 0-9. |
0.83104 |
FastText skip-gram |
Spell out digits 0-9. |
0.83688 |
FastText skip-gram |
Spell out digits 0-10. |
0.8323 |
FastText skip-gram |
Spell out digits 0-9. Supervised on pretrained vectors cc.en.300. |
0.8290 |
FastText skip-gram |
Spell out digits 0-10. Supervised on pretrained vectors wiki-news-300d-1M. |
0.8276 |
FastText skip-gram |
Spell out digits 0-9. Supervised on pretrained wiki-news-300d-1M. |
0.8260 |
FastText skip-gram |
Spell out digits 0-9. Directly uses pretrained vectors cc.en.300. |
0.7582 |
FastText cbow |
Preserve stop words, numbers. |
0.80956 |
"Supervised on pretrained vectors" means initialize the model with pretrained vectors and train it on the data set.
Word Embedding |
Unit |
Dropout |
Epoch |
Details |
Score |
Word2Vec skip-gram |
32 |
0.2 |
6 |
Spell out digits 0-9. |
0.87232 |
Word Embedding |
Unit |
Dropout |
Epoch |
Details |
Score |
Word2Vec skip-gram |
32 each direction |
0.2 |
6 |
Spell out digits 0-9. |
0.88024 |
Word Embedding |
CNN num_filters |
CNN kernel_size |
LSTM Unit |
Epoch |
Details |
Score |
Word2Vec skip-gram |
64 |
2 |
32 |
6 |
Spell out digits 0-9. |
0.89948 |
FastText skip-gram |
64 |
2 |
32 |
6 |
Spell out digits 0-9. |
0.90132 |