Skip to content
master
Switch branches/tags
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Pivot-based Entity Linking

Pivot-based entity linking (PBEL) is an entity linking method that uses high-resource languages as intermediates while doing low-resource cross-lingual entity linking. The details are described in the paper: Zero-shot Neural Transfer for Cross-lingual Entity Linking.

Dependencies

Data

  • Download sample data for one language pair here.
  • Download data for all 54 training languages and 9 test languages here.

Usage

  • train.py is used for training the entity similarity model.
    • Default values for hyperparameters are in the script -- 64 size character embeddings and 1024 hidden size for the character LSTM.
  • test.py is used both for encoding the knowledge base as well as retrieving entity linking candidates for test data or other input files.
  • See the scripts/ folder for examples using data from here.

References

If you use this repository, please cite

@inproceedings{rijhwani19aaai,
    title = {Zero-shot Neural Transfer for Cross-lingual Entity Linking},
    author = {Shruti Rijhwani and Jiateng Xie and Graham Neubig and Jaime Carbonell},
    booktitle = {Thirty-Third AAAI Conference on Artificial Intelligence (AAAI)},
    address = {Honolulu, Hawaii},
    month = {January},
    url = {https://arxiv.org/abs/1811.04154},
    year = {2019}
}

About

Code and data for pivot-based entity linking.

Resources

Releases

No releases published

Packages

No packages published