• on MacOS system

$ git clone https://github.com/gnodisnait/mushroom.git
$ cd mushroom
$ virtualenv -p python3 venv
$ source venv/bin/activate
(venv) $ pip install -r requirements_mac.txt

• on Ubuntu system

$ git clone https://github.com/gnodisnait/mushroom.git
$ cd mushroom
$ virtualenv -p python3 venv
$ source venv/bin/activate
(venv) $ pip install -r requirements_ubuntu.txt

open mushroom/config.py file, set dataPath to the absolute path of your datasets for this project; download datasets

(venv) $ python mushroom.py --func download

• If successful, three sub-directories with datasets at dataPath directory will be created: Freebase13, Wordnet11, and Wordnet18

(venv) $ python mushroom.py --func tc --kg [Wordnet18|Wordnet11|Freebase13|all] --e2v [TH|TE|TransE|all]

To run all the triple classification tasks on all knowledge-graphs, with different pre-trained entity-embeddings, type

(venv) $ python mushroom.py --func tc --kg  all --e2v all

It takes more than 14 hours to finish running the above command

(venv) $ python mushroom.py --vis_gamma [precision|recall|accuracy]  --kg [Wordnet18|Wordnet11|Freebase13]

(venv) $ python mushroom.py --vis_length [precision|recall|accuracy]  --kg [Wordnet18|Wordnet11|Freebase13]

(venv) $ python mushroom.py --vis_mg [precision|recall|accuracy] --kg [Wordnet18|Wordnet11|Freebase13]  --legendLoc lower right

If you use the code, please cite the following paper:

Tiansi Dong, Zhigang Wang, Juanzi Li, Christian Bauckhage, Armin B. Cremers (2019). Triple Classification Using Regions and Fine-Grained Entity Typing. AAAI-19 The Thirty-Third AAAI Conference on Artificial Intelligence, January 27 – February 1, 2019 Hilton Hawaiian Village, Honolulu, Hawaii, USA.

Zhigang Wang, Juanzi Li (2016). Text-Enhanced Representation Learning for Knowledge Graph. IJCAI-16 July 9 -- 16, 2016 New York, USA.