Reasoning With Neural Tensor Networks for Knowledge Base Completion

An expressive neural tensor network suitable for reasoning over relationships between two entities given a subset of the knowledge base is introduced and performance can be improved when entities are represented as an average of their constituting word vectors.