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