classification

Dear all,

I have done a refined named entity recognition (NER), I have found coordinate, date, time,...however, the coordinate has a minor problem. NOW based on these labels I gave to each sentence a label (1-->obsedrvational sentences) mean that this sentence is related to observation, I want to know how can I proceed with prodigy to classify sentences or make annotation edit to provide more accurate training data, I already used my ml and dl method to classification, but for one class (1) the classication result after work it is not very good?

precision    recall  f1-score   support

           0       0.94      1.00      0.97      1555
           1       0.80      0.20      0.32       120

    accuracy                           0.94      1675
   macro avg       0.87      0.60      0.64      1675
weighted avg       0.93      0.94      0.92      1675

can you give me some hints regarding prodigy and also ML, do you think it is a good direction?

Hi,

I'm having a bit of trouble understanding your question. From what I understand, you're looking for ways to improve the accuracy of your model?

There's not really a good general-purpose answer to that, and many of the issues go outside the scope of Prodigy. The annotation scheme, annotation consistency and number of annotations all play a big role. You could also experiment with using the data you've labelled with different tools, instead of just with spaCy, which powers Prodigy's default models.

1 Like