No description, website, or topics provided.
Switch branches/tags
Nothing to show
Clone or download

ELDEN: Improved Entity Linking using Densified Knowledge Graphs

This software is the implementation of the paper "ELDEN: Improved Entity Linking using Densified Knowledge Graphs" to be presented at 16th Annual Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL HLT 2018) at New Orleans, Louisiana, June 1 to June 6, 2018.


Code is written in Python (2.7), Torch and Lua (Luajit)

Using the pre-trained word2vec vectors from gensim will require downloading it from

Co-occurance matrix and other datafiles can be downloaded at

Running the models

This package contains the four steps (folders A to D) of implementation, followed by Evaluation. We suggest running the system in this order.

A. Corpus :

  1. Wikipedia (clean as specified in paper)
  2. Web Corpus =,,

B. Dataset :

  1. TAC2010 = TACforNED
  2. CoNLL = Please cite the respective papers when using these datasets.

C. Preprocess:

  1. Create entity co-location index. python2.7 base_co.npy/None vocab.pickle output_file file_scraped_from_web
  2. Start PMI Server. python
  3. Train entity embeddings. th> main.lua <<word2vec.lua>>
  4. Start Embedding Distance Servers. th> EDServer.lua

D. Entity Linker:

  1. Create train and test dataset python
  2. Run Entity Linker python

E. Evaluation :

  1. Head entities versus tail entities statistics python

Kindly cite the paper if you are using the software