KGRefiner

KGRefiner is a novel refining method for refining knowledge graphs. This method helps link prediction algorithms outperform by using hierarchical information in knowledge graphs. Paper: KGRefiner: Knowledge Graph Refinement for Improving Accuracy of Translational Link Prediction Methods.

Inferable Persian Knowledge Graph

Graph I worked on the Persian knowledge graph. The original graph was not inferable; therefore, no link prediction and triple classification methods could infer on them. I developed several algorithms to prune the graphs. As a result, Farsbase107k392R is created from Farsbase. (comming soon)

Hadware Limited OpenKE

OpenKE is a widely used framework for link prediction. It does not work on Google Colab due its time execution limits. I made pause and continue script for OpenKE so everyone can extract knowledge even on big graphs with Colab.The repository is on Github.

Knowledge Graph Generator

I was a part of a project related to generating a knowledge graph from the raw Persian text in Shahid Beheshti University. Since I am experienced with knowledge graphs, I proposed a way of generating graphs by NER and other existing graphs.

Predicting the Popularity of News Topics on the Tabnak News Website

As my bachelor’s final thesis, I developed a software to predict the popularity of different news categories. This web-based software extracts various news from a news website every day and shows users the number of changes in views and popularity of topics in the form of a dashboard.