NASARI semantic vector representations for BabelNet synsets* and Wikipedia pages in several languages. Currently available three vector types: lexical, unified and embedded. NASARI provides a large coverage of concepts and named entities and has been proved to be useful for many Natural Language Processing tasks such as multilingual semantic similarity, sense clustering or word sense disambiguation, tasks on which NASARI has contributed to achieve state-of-the-art results on standard benchmarks.
*Please note that BabelNet covers WordNet and Wikipedia among other resources, enabling our vectors to be applicable for representations of concepts and named entities in each of these resources.
Downloads
Currently available for English, Spanish, French, German and Italian. Please find more information in the README file.
Stay tuned for the release of NASARI representations in other languages!
The NASARI embed vectors below share the same space with the pre-trained vectors of Word2Vec for English (trained on the Google News corpus),
or with the Word2Vec word embeddings trained on the Spanish Billion Words Corpus for Spanish (more information in the main reference paper).
You can download the Spanish word embeddings here.
New (July 2017): Now you can additionally download the 300-dimensional NASARI-embed concept and entity BabelNet synset embeddings along with the Word2Vec word embeddings trained on the UMBC corpus, both in the same vector space. These vectors tend to show a superior performance than the NASARI-embed vectors trained on Google News below. Download both NASARI-embed and the UMBC word embeddings here (note that in this version all word embeddings are lowercased):
-
NASARI-embed and UMBC word embeddings in a single compressed bin file (compatible with Word2Vec and gensim):
[7GB]
-
NASARI-embed and UMBC word embeddings in txt from a compressed zip file: [8GB]
Note: the first three lines of the table below correspond to the NASARI vector representations for all English Wikipedia pages (Wikipedia dump of November 2014). In the remaining files each vector is tagged with its corresponding BabelNet synset and Wikipedia page.
| Language | Type | # of BabelNet synsets | # of Wikipedia pages | Download | Size |
| English | Lexical(Wiki) | - | 4.40M | |
4.7GB |
| English | Embed(Wiki) | - | 4.40M | |
5.9GB |
| English | Unified(Wiki) | - | 2.85M | |
341MB |
| English | Lexical | 4.42M | 4.40M | |
4.7GB |
| English | Embed | 4.42M | 4.40M | |
5.9GB |
| English | Unified | 2.87M | 2.85M | |
352MB |
| Spanish | Lexical | 1.07M | 1.05M | |
705MB |
| Spanish | Unified | 657K | 635K | |
60MB |
| Spanish | Embed | 1.07M | 1.05M | |
1.4GB |
| French | Lexical | 1.48M | 1.45M | |
1.1GB |
| French | Unified | 882K | 861K | |
96MB |
| German | Lexical | 1.51M | 1.49M | |
1.4GB |
| German | Unified | 857K | 836K | |
59MB |
| Italian | Lexical | 1.10M | 1.08M | |
843MB |
| Italian | Unified | 631K | 610K | |
69MB |
The NASARI_embed vector representations can also be downloaded in binary format: [bin:4.8GB] (compatible with Word2Vec).
NASARI lexical vectors in English can also be downloaded in tar.bz2 compression format: [tar.bz2:3.6GB].
Please note that you can use the BabelNet API to get the most from these vectors, e.g., access the corresponding WordNet synsets or lexicalizations.
Release history
| Release version | Date | Features | Reference paper |
| 1.0 | April 2015 | English lexical and unified vectors for WordNet synsets and Wikipedia pages. | NAACL 2015 |
| 2.0 | August 2015 | Multilingual extension through BabelNet. Available in English, Spanish, French, German and Italian. | ACL 2015 |
| 2.1 | October 2015 | Minor bug fixing and updated format. | - |
| 3.0 | March 2016 | Improved lexical and unified vectors. Integration of embedding vector representations. | AIJ 2016 |
Main reference
If you use any of the resources available in this website, please refer to the following article [bib]:
José Camacho-Collados, Mohammad Taher Pilehvar
and Roberto Navigli.
Nasari: Integrating explicit knowledge and corpus statistics for a multilingual representation of concepts and entities.
Artificial Intelligence 240, Elsevier, 2016, pp.567-577.
@article{camacho2016nasari,
title={Nasari: Integrating explicit knowledge and corpus statistics for a multilingual representation of concepts and entities},
author={Camacho-Collados, Jos{\'e} and Pilehvar, Mohammad Taher and Navigli, Roberto},
journal={Artificial Intelligence},
volume={240},
pages={36--64},
year={2016},
publisher={Elsevier}
}
Previous reference papers
José Camacho-Collados, Mohammad Taher Pilehvar and Roberto Navigli.NASARI: a Novel Approach to a Semantically-Aware Representation of Items.
In Proceedings of the North American Chapter of the Association of Computational Linguistics (NAACL 2015), Denver, USA, pp. 567-577, 2015
José Camacho-Collados, Mohammad Taher Pilehvar and Roberto Navigli.
A Unified Multilingual Semantic Representation of Concepts.
In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics (ACL 2015), Beijing, China, July 27-29, pp. 741-751, 2015
Contact
Should you have any enquiries about any of the resources, please contact us:
