Named Entity Recognition (NER)

Named entity recognition is the task of identifying named entities like person, location, organization, drug, time, clinical procedure, biological protein, etc. NER systems are often used as the first step in question answering, information retrieval, co-reference resolution, topic modeling etc because they help in narrowing down the important parts of the text.

Evaluation Metrics

The most common metric used for NER if F1.

Models

The intialize models were based on hand crafted rules. I am at <location> etc to bootstap the model ones we have a intialize set of entites we can mine more rules based on them. As NER is classifcation task classical ML algorithms like SVM, CRF etc were also used here a big importance is given to the feature engineering aspect. The models like RNN, LSTM, Bi-LSTM, CNN, Bi-LSTM-CRF, etc were used to solve this task. After the advent of transformers models like BERT, RoBERTa, etc were used to solve this task and they have achieved state-of-the-art results but they are computationally expensive and the improvements vs the cost of operation depends on the use case need to be considered before using them.