Skip to content

Latest commit

 

History

History
19 lines (10 loc) · 968 Bytes

File metadata and controls

19 lines (10 loc) · 968 Bytes

#Path Query experiments

This is the path query experiment in Sec 5.5 of the paper.

##Data Download the dataset released by the original paper here

For evaluation, you will also need to have the negative entities for each target entity. You can get that by running data/get_negative_examples.py. (You have to run it in the original code repository.). I am unable to share the negative entities because of the data limit of github repositories. Link to negative samples used for experiments.

##Running the model Checkout model/run.lua. It defines all the hyper-params and sets other required variables. If you want to train the model on a CPU instead of a GPU just set useCuda = False. To start training the model

th run.lua