The key task of NER is to identify and classify entities from a text into pre-defined categories like person names, locations, organizations, dates, quantities and many more.

NER works on 2 stages:

1. Identification: first find out the boundary of the current entity (start, end)
2. Classification: then classify the current selected boundary into one of the categories

Methods:

1. Rule based approaches: using patterns or dictionaries
2. Statistical models: Hidden Markov Model, Conditional Random Field
3. Neural network based methods: Transformer, RNN, LSTM, BERT

Applications:

1. Information extraction
2. Customer report generation
3. Search engine
4. News aggregation

References