Vietnamese Semantic Role Labelling (vnSRL) v1.0.0 - 2016-02-20

---

Code by Thai-Hoang Pham, Xuan-Khoai Pham

1. Introduction

The vnSRL system is used to labelling semantic roles of arguments for each predicate in a Vietnamese sentence. This software is written by Python 2.x.

2. Installation

This software depends on NumPy, SciPy, Scikit-learn, Pandas, Pulp, ETE2, six Python packages for scientific computing. You must have them installed prior to using vnSRL.

The simple way to install them is using pip:

	# pip install -U numpy scipy scikit-learn pandas pulp ete2

3. Usage

3.1. Data

The input data's format of vnSRL is Penn Treebank format. For details, see sample data 100-sen.txt in a directory 'data/input'.

For classifying task, put your data in a directory 'data/input'.

Your output file is stored in a directory 'data/output'.

3.2. Command-line Usage

You can use vnSRL software by a following command:

	$ python vnSRL.py <input> <ilp> <embedding> <output>

Positional arguments:

• input: input file name
• ilp: integer linear programming post-processing (1 for using ilp and 0 for vice versa)
• embedding: word embedding file (skipgram or glove)
• output: output file name

For example, if you want to use this software to label the file input.txt, ilp for post-processing, glove for word embedding and the output file output.txt, you use the command:

	$ python vnSRL.py input.txt 1 glove output.txt

Note: In our experiment, integer linear programming method helps to improve the performance about 0.4% but takes very long time to run (about 50x longer).

4. References

T-H. Pham, X-K. Pham, and P. Le-Hong, "Building a Semantic Role Labelling System for Vietnamese", Proceedings of the 10th International Conference on Digital Information Management, Jeju Islands, South Korea, IEEE, 10/2015.

5. Contact

Thai-Hoang Pham < [email protected] >

FPT Technology Research Institute, FPT University