Description
Multi labels text classification
Models
Models are convolutional neural network with a word embedding layer custom trained with domain specific corpus. models are individually trained per y_
Here is dataset sample
Here is train/test validation report sample. Validation accuracy scores are bounded between 40~60%
Training Category-Classifier Now...
Train on 3288 samples, validate on 823 samples
Epoch 10/20
3288/3288 [==============================] - 91s - loss: 1.3141 - acc: 0.5289 - val_loss: 1.3562 - val_acc: 0.5431
Training Component-Classifier Now...
Train on 3288 samples, validate on 823 samples
Epoch 1/20
3288/3288 [==============================] - 101s - loss: 3.1712 - acc: 0.2205 - val_loss: 2.9410 - val_acc: 0.1324
Repo Structure
- classifiers -> models + training / update
- crawler -> crawling runtime
- data-service -> api service
- web -> ui