Custom Text Classification Flashcards
Which are the two types of labelling solution
SingleLable
Multi-Lable
what is meant by classification
To assign a class to an object, text or image or element within an image. The classification itself is akin to how we think of a movie genre to classify a move as thriller, adventure, fantasy…
In Machine Learning parlance, it is to assign lables to text or other information. This is done to prepare a training set to train a model which is then consumed to identify the classification of similar object.
Three metrics used to evaluate a trained model performance
Recall - % of predited lable out of the total test data
Precision - accuracy of the result
F1 Score is the normalised overall score taking into account both recall and precision.
Two modes of splitting data
Manual
Automatic split
AI Language model lifecycle
Define labels
Tag data
train model
view model performance
improve model
deploy model
classify text
What is the API pattern for text classification
Asynchronous. we submit initial result, followed by another call to get result status by using the URL and object id provided in response to the initial call