Code for the EMNLP paper, "The Importance of Joint Multilingual Supervision for Cross-Lingual Entity Linking".

Preprocessing Wikipedia for Candidate Generation

First, preprocess the Wikipedia dump for the languages you care about using the wikidump_preprocessing code here.

Setting up Candidate Generation

The code for performing candidate generation is available at wiki_candgen. It uses the resources generated using the wikidump_preprocessing repo.

Model

Code for the model is available at xling-el. Follow instructions there.

@inproceedings{UGR18,
  author = {Upadhyay, Shyam and Gupta, Nitish and Roth, Dan},
  title = {Joint Multilingual Supervision for Cross-lingual Entity Linking},
  booktitle = {EMNLP},
  year = {2018}
}```