Nested-NER

Nested-NER is an implementation of [A Neural Layered Model for Nested Named Entity Recognition] (http://aclweb.org/anthology/N18-1131).

Requirements

• Ubuntu 16.04
• chainer 3.3.0
• python 3.5.2
• numpy 1.14.1
• cupy 2.4.0
• cuda 9.1
• cudnn 7.0

Data format

Each line has multiple columns separated by a tab key. Each line contains

word	label1	label2	label3	...	labelN

The number of labels (N) for each word is determined by the maximum nested level in the data set. N=maximum nested level + 1 Each sentence is separated by an empty line. For example, for these two sentences, John killed Mary's husband. He was arrested last night , they contain four entities: John (PER), Mary(PER), Mary's husband(PER),He (PER). The format for these two sentences is listed as following:

John    B-PER   O   O
killed  O   O   O
Mary    B-PER   B-PER   O
's  O   I-PER   O
husband O   I-PER   O
.   O   O   O

He    B-PER   O   O
was  O   O   O
arrested  O   O   O
last  O   O   O
night  O   O   O
.  O   O   O

Pretrained word embeddings

• Pretrained word embeddings used in GENIA: https://drive.google.com/open?id=0BzMCqpcgEJgiUWs0ZnU0NlFTam8
• Pretrained word embeddings used in ACE2005:http://tti-coin.jp/data/wikipedia200.bin

Configuration

Parameters are listed in the config file which is located in the layered-bilstm-crf/src folder. Before running the codes, please change the parameters with specific values.

Usage

Training

cd layered-bilstm-crf/src/
python3 train.py

Testing

cd layered-bilstm-crf/src
python3 test.py

Please cite our NAACL paper when using this code.

• Meizhi Ju, Makoto Miwa, Sophia Ananiadou. A Neural Layered Model for Nested Named Entity Recognition In the Proceedings of NAACL-HLT2018, pp. 1446--1459.