persian named entity reognition using lstm-crf. both character and word embeddings are fed into lstm and is trained for classification task.
finally by feeding new sentence to network, All named entities related to words in the sentence will be predicted.
accuracy was almost 90 percent on test data.