IMDb30 is a film KG dataset, formed as triplet (h, r, t).

Paper: IMDb30: A Multi-relational Knowledge Graph Dataset of IMDb Movies

The code for constructing IMDb30 is in imdb_movie_KG/imdb_movie_KG.ipynb, the original triplets are stored in imdb_movie_KG/IMDB30, the divided dataset that can be used for model training and evaluating are stored in imdb_movie_KG/IMDB_benchmark.

The code for convolution model (ConvKB, ConvE, Conv3D) training and evaluating is in model_4_imdb.

The code for embedding visualization is in visualization.