This is a brief summary of paper for me to study and arrange for Named Entity Recognition with Bidirectional LSTM-CNNs (Chiu and Nichols., TACL 2016) I read and studied.

This paper is a research ralted to NER tagging.

They designed the model with Bi-LSTM-CNN. The CNN extracted the information of characters consisting of the word.

Their model is as follows:

The whole model is

To extract a feature vector from the per-character vector

They used the BIOES annotation to label the word into the corresponding catergorical tag, methioning this scheme has reported to outperform others such as BIO(Ratino and Roth, 2009 CoNLL)

• BIOES annotation (Begin, Inside, Outside, End, Single)

based on their experiement, they said that the plain SGD is bettern than other optimization algorithms.

their dataset preprocessing:

• All digit sequences are replaced by a single “0”.
• Before training, we group sentences by word length into mini-batches and shuffle them.

Reference

• Paper
  • arXiv Version: Named Entity Recognition with Bidirectional LSTM-CNNs (Chiu and Nichols., arXiv 2015)
  • TACL Version: Named Entity Recognition with Bidirectional LSTM-CNNs (Chiu and Nichols., TACL 2016)
• How to use html for alert
  • how to use icon