blob: c67ec0ec95ae591b974b4e9de3574fc8a3f98762 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
|
# Evaluating cross-lingual textual similarity on dictionary alignment
This repository contains the scripts to prepare the resources as well as open source implementations of the methods. Word Mover's Distance and Sinkhorn implementations are extended from [Cross-lingual retrieval with Wasserstein distance](https://github.com/balikasg/WassersteinRetrieval).
## Requirements
```bash
pip install -r requirements.txt
```
- Python 3
- [nltk](http://www.nltk.org/)
```python
import nltk
nltk.download('wordnet')
```
- [lapjv](https://pypi.org/project/lapjv/)
- [POT](https://pypi.org/project/POT/)
- [mosestokenizer](https://pypi.org/project/mosestokenizer/)
- (Optional) If using VecMap
* NumPy
* SciPy
<details><summary>We recommend using a virtual environment</summary>
<p>
In order to create a [virtual environment](https://docs.python.org/3/library/venv.html#venv-def) that resides in a directory `.env` under home;
```bash
cd ~
mkdir -p .env && cd .env
python -m venv evaluating
source ~/.env/evaluating/bin/activate
```
Inside the virtual environment, the python interpreter and the installed packages are isolated.
In order to install all dependencies automatically use the [pip](https://pypi.org/project/pip/) package installer using `requirements.txt`, which resides under the repository directory.
```bash
# under Evaluating-Dictionary-Alignment
pip install -r requirements.txt
```
Rest of this README assumes that you are in the repository root directory.
</p>
</details>
## Acquiring The Data
```bash
git clone https://github.com/yigitsever/Evaluating-Dictionary-Alignment.git && cd Evaluating-Dictionary-Alignment
./get_data.sh
```
This will create two directories; `dictionaries` and `wordnets`.
Linewise aligned definition files are in `wordnets/ready`.
## Acquiring The Embeddings
We use [VecMap](https://github.com/artetxem/vecmap) on [fastText](https://fasttext.cc/) embeddings. You can skip this step if you are providing your own polylingual embeddings.
Otherwise;
* initialize and update the VecMap submodule;
```bash
git submodule init && git submodule update
```
* make sure `./get_data` is already run and `dictionaries` directory is present.
* run;
```bash
./get_embeddings.sh
```
Bear in mind that this will require around 50 GB free space.
|
