This repo is DEPRECATED and is not supported anymore.
This repository contains implementations for a variety of AKBC (Automatic Knowledge Base Completion) scoring models. In its current version only scoring models for triples (subject, predicate, object) are supported.
It also supports compositional scoring models using various composition functions.
Currently only the FB15k-237 dataset is supported, which was introduced here and here.
For training see command line options of train.py.
requires: tensorflow, pandas
For further questions, please don't hesitate to contact.
Contact: dirk.weissenborn@dfki.de