Leveraging Lexical Resources for Learning Entity Embeddings in Multi-Relational Data
This work addresses the challenge of improving relational models for knowledge base completion, though it is incremental as it builds on existing methods like TransE.
The authors tackled the problem of learning vector-space embeddings for multi-relational data by proposing a method that leverages lexical resources for better initialization, resulting in a new state-of-the-art mean rank of 51 on the WordNet dataset, down from 212, and faster convergence.
Recent work in learning vector-space embeddings for multi-relational data has focused on combining relational information derived from knowledge bases with distributional information derived from large text corpora. We propose a simple approach that leverages the descriptions of entities or phrases available in lexical resources, in conjunction with distributional semantics, in order to derive a better initialization for training relational models. Applying this initialization to the TransE model results in significant new state-of-the-art performances on the WordNet dataset, decreasing the mean rank from the previous best of 212 to 51. It also results in faster convergence of the entity representations. We find that there is a trade-off between improving the mean rank and the hits@10 with this approach. This illustrates that much remains to be understood regarding performance improvements in relational models.