Efficient Entity Embedding Learning with StarSpace

Github repositories

• facebookresearch/StarSpace

  StarSpace is a versatile neural model that enables efficient learning of entity embeddings for various tasks. It can learn word, sentence, or document level embeddings and supports applications such as information retrieval, text classification, recommendation systems, and image analysis. The model represents different types of objects in a shared vectorial embedding space, allowing for comparisons and ranking. StarSpace can be built on Mac OS, Windows, or Linux and requires a compiler with good C++11 support. It provides a flexible file format for input data and offers different training modes depending on the specific task.