In [1]:

    

from process_data import DataHandler


    

    




Using TensorFlow backend.

In [19]:

    

s = DataHandler("../data/CoNLL-2003/eng.testa")


    

In [3]:

    

s.get_data()[0].shape


    

    
Out[3]:





(3466, 109, 60)

In [4]:

    

from NER_model import NER


    

In [5]:

    

m = NER(s)


    

In [6]:

    

m.make_and_compile()


    

    




____________________________________________________________________________________________________
Layer (type)                     Output Shape          Param #     Connected to                     
====================================================================================================
bidirectional_1 (Bidirectional)  (None, 109, 300)      253200      bidirectional_input_1[0][0]      
____________________________________________________________________________________________________
timedistributed_1 (TimeDistribut (None, 109, 5)        1505        bidirectional_1[0][0]            
____________________________________________________________________________________________________
dropout_1 (Dropout)              (None, 109, 5)        0           timedistributed_1[0][0]          
====================================================================================================
Total params: 254,705
Trainable params: 254,705
Non-trainable params: 0
____________________________________________________________________________________________________
None

In [ ]:

    

# m.train()
m.train(epochs=10)


    

In [12]:

    

m.evaluate()


    

    




             precision    recall  f1-score   support

        LOC       0.91      0.71      0.80       471
       MISC       0.77      0.47      0.59       276
          O       0.00      0.00      0.00         0
        ORG       0.80      0.49      0.61       362
        PER       0.92      0.81      0.86       635

avg / total       0.87      0.66      0.75      1744







    
Out[12]:






  

    

      

      
LOC
      
MISC
      
O
      
ORG
      
PER
    
  
  

    

      
LOC
      
334
      
23
      
81
      
26
      
7
    
    

      
MISC
      
15
      
130
      
117
      
9
      
5
    
    

      
O
      
0
      
0
      
0
      
0
      
0
    
    

      
ORG
      
17
      
13
      
124
      
177
      
31
    
    

      
PER
      
0
      
2
      
113
      
8
      
512
    
  

In [17]:

    

# m.predict_tags("The strongest man on Earth is Mark Henry")
m.predict_tags("The strongest rain ever recorded in India shut down the financial hub of Mumbai, snapped communication lines, closed airports and forced thousands of people to sleep in their offices or walk home during the night, officials said today.")


    

    




(1, 109, 60)
The/O
strongest/O
rain/O
ever/O
recorded/O
in/O
India/LOC
shut/O
down/O
the/O
financial/O
hub/O
of/O
Mumbai,/LOC
snapped/O
communication/O
lines,/O
closed/O
airports/O
and/O
forced/O
thousands/O
of/O
people/O
to/O
sleep/O
in/O
their/O
offices/O
or/O
walk/O
home/O
during/O
the/O
night,/O
officials/O
said/O
today./O

In [ ]:

    

m.model.save("./first_model")


    

In [8]:

    

from keras.models import load_model
m.model = load_model("./first_model")