Which Deep Learning technique is used in NER and why?

Many deep learning techniques are used in NER, some of which are very popular and there is a reason behind that. One such technique is LSTM technique. LSTM technique is suitable because of the fact that the text is considered as a sequential data format. Since LSTM only takes the past information from the sequence into consideration, Bi-directional LSTM is preferred because named entities require both past and future labels in a sequence.

The following architecture shows how bi-directional LSTM works.