https://github.com/ozanarkancan/char-ner
Multi lingual character based named entity recognizer
https://github.com/ozanarkancan/char-ner
bidirectional-lstms character-level deep-learning multilingual named-entity-recognition
Last synced: 10 months ago
JSON representation
Multi lingual character based named entity recognizer
- Host: GitHub
- URL: https://github.com/ozanarkancan/char-ner
- Owner: ozanarkancan
- Created: 2015-07-06T08:57:42.000Z (about 11 years ago)
- Default Branch: master
- Last Pushed: 2018-04-08T13:56:47.000Z (over 8 years ago)
- Last Synced: 2025-04-19T18:50:24.258Z (about 1 year ago)
- Topics: bidirectional-lstms, character-level, deep-learning, multilingual, named-entity-recognition
- Language: Python
- Homepage:
- Size: 333 MB
- Stars: 25
- Watchers: 7
- Forks: 8
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
## CharNER
For more details, please refer to the paper: [CharNER: Character-Level Named Entity Recognition](http://www.aclweb.org/anthology/C/C16/C16-1087.pdf)
## Usage
Access to full list of options by typing:
```
python exper.py --help
```
Example command:
```
python src/exper.py --activation bi-lstm --n_hidden 128 128 --drates .2 .5 .8 --lang cze
```
This command builds 2 Bidirectional LSTMs stacked of top of each other.
Each forward and backward LSTM has 128 units.
--drates (dropout rates) flag signals to use dropout.
In this example, .2 dropout is applied to inputs (drops characters) and .5 & .8 dropouts are applied to the outputs of Bidirectional LSTMs.
--lang flag dictates which folder to use under data/ directory
## Data Format
Each folder under data/ directory is composed of 3 files.
train.bio, testa.bio, testb.bio are for training, development and test sets respectively.
Each file contains word, tag pairs seperated with a tab.
For examples, check out directories under data/.
## Install Dependencies
```
pip install --upgrade https://github.com/Lasagne/Lasagne/archive/master.zip
pip install -r requirements.txt
```
## Authors
* Onur Kuru
* Ozan Arkan Can
## Miscellaneous
You can checkout the Keras version [here](https://github.com/0xnurl/keras_character_based_ner).