LibN3L: A light-weight neural network package for natural language.

Just include the directory in your code and call it by "#include N3L.h"

Installation:

Prerequisition:
    mshadow
Please download and include the directory "https://github.com/dmlc/mshadow/tree/master/mshadow" in your applications:
(a) first copy the directory into your computer;
(b) then include it in your applications;
If have any problems, please mason.zms@gmail.com

If you want to make mshadow work, you need to install certain libaries such as openblas and cuda.
I suggest use openblas since the current version does not support cuda yet, which is our future work.
Find it here:
https://github.com/xianyi/OpenBLAS
Compile and install:
make USE_THREAD=0     ##single thread version, one can use multi-thread version as well.
make install            ##default path /opt/OpenBLAS
cp /opt/OpenBLAS/include/. /usr/include/
cp /opt/OpenBLAS/lib/. /usr/lib(64)/

Examples:

Some examples are realeased at:
https://github.com/SUTDNLP/NNSegmentation
https://github.com/SUTDNLP/NNPOSTagging
https://github.com/SUTDNLP/NNNamedEntity (already support model save/load module)
You can see the performances in [description.pdf](description(expect for lrec2016).pdf)
https://github.com/SUTDNLP/OpenTargetedSentiment

Cite:

If you use Libn3l for research, please cite our LREC paper as follows:

@article{zhang2016libn3l,
title={Libn3l: a lightweight package for neural nlp},
author={Zhang, Meishan and Yang, Jie and Teng, Zhiyang and Zhang, Yue},
booktitle={Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)},
pages={225--229},
year={2016} 
} 

Updating...

• 2015-12-10: support model saving and loading.
• 2015-10-20: initial version is available at GitHub.
• 2015-5-31: framework is completed.
• 2014-12-01: project starts.