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
.