Wikipedia2Vec is a tool used for obtaining embeddings (or vector representations) of words and entities (i.e., concepts that have corresponding pages in Wikipedia) from Wikipedia. It is developed and maintained by Studio Ousia.

This tool enables you to learn embeddings of words and entities simultaneously, and places similar words and entities close to one another in a continuous vector space. Embeddings can be easily trained by a single command with a publicly available Wikipedia dump as input.

This tool implements the conventional skip-gram model to learn the embeddings of words, and its extension proposed in Yamada et al. (2016) to learn the embeddings of entities. This tool has been used in several state-of-the-art NLP models such as entity linking, named entity recognition, entity relatedness, and question answering.

This tool has been tested on Linux, Windows, and macOS.

An empirical comparison between Wikipedia2Vec and existing embedding tools (i.e., FastText, Gensim, RDF2Vec, and Wiki2vec) is available here.

Pretrained embeddings for 12 languages (i.e., English, Arabic, Chinese, Dutch, French, German, Italian, Japanese, Polish, Portuguese, Russian, and Spanish) can be downloaded from this page.


If you use Wikipedia2Vec in a scientific publication, please cite the following paper:

Ikuya Yamada, Akari Asai, Hiroyuki Shindo, Hideaki Takeda, Yoshiyasu Takefuji, Wikipedia2Vec: An Optimized Tool for Learning Embeddings of Words and Entities from Wikipedia.

  title={Wikipedia2Vec: An Optimized Tool for Learning Embeddings of Words and Entities from Wikipedia},
  author={Yamada, Ikuya and Asai, Akari and Shindo, Hiroyuki and Takeda, Hideaki and Takefuji, Yoshiyasu},
  journal={arXiv preprint 1812.06280},


Apache License 2.0