https://github.com/thunlp/COS960

COS960: A Chinese Word Similarity Dataset of 960 Word Pairs
https://github.com/thunlp/COS960

Last synced: 6 days ago
JSON representation

COS960: A Chinese Word Similarity Dataset of 960 Word Pairs

• Host: GitHub
• URL: https://github.com/thunlp/COS960
• Owner: thunlp
• License: mit
• Created: 2019-06-01T07:15:23.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-06-06T11:20:21.000Z (over 5 years ago)
• Last Synced: 2024-08-02T05:05:49.495Z (4 months ago)
• Language: Python
• Size: 20.5 KB
• Stars: 36
• Watchers: 9
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-hackchinese - C0S960

README

        
# COS960

COS960 is a Chinese word similarity dataset of 960 word pairs. Each pair of words is annotated by  15 native speakers with a similarity score which reflects **true similarity**. The 960 word pairs are further divided into 3 groups according to their Part Of Speech tags, including 480 pairs of nouns, 240 pairs of verbs and 240 pairs of adjectives.

### Usage

To use COS960 to test your word embedding, use command

```

python correlation_calcu.py {VECTOR_FILE}

```

### Dataset

The data in the files is formulated as

```

[Word1] [Word2] [Average] [Annotator1] ... [Annotator15]

小心谨慎  谨慎小心     4.0         4      ...       4 

```

### Cite

If you  use the dataset, please cite this:

```

@article{huang2019COS960,

Author = {Junjie Huang and Fanchao Qi and Chenghao Yang and Zhiyuan Liu and Maosong Sun},

Title = {{COS960: A Chinese Word Similarity Dataset of 960 Word Pairs}},

journal={arXiv preprint arXiv:1906.00247},

Year = {2019},

}

```