YAGO2: A Spatially and Temporally Enhanced Knowledge Base from Wikipedia

Research Demo Downloads Evaluation Publications People Applications

If you are using the YAGO(2) ontology, please let us know and send a mail to NameOfTheOntology@mpi-inf.mpg.de!
We would really like to know who is using YAGO, and how you are using it. We'd also like to put a link to you on our applications page as a user of YAGO, if you are ok with that.

YAGO2 Downloads

The YAGO2 ontology is based on data from Wikipedia, WordNet and GeoNames, and includes the translations of classes into many languages from Universal WordNet (UWN). We make available two versions of YAGO2: core and full.

Release Notes

2012-01-09

Core

Core contains all the entities and ontological facts extracted from Wikipedia, with categories mapped to WordNet classes. In addition, we include facts extracted from GeoNames for all the entities we can map to the Wikipedia ones (i.e. we extract only facts for the entities that were extracted from Wikipedia, and not add entities only present in GeoNames). It also contains multi-lingual data from UWN. In total, it contains 2.6 million entities and about 33 million facts.

YAGO2 core in the default YAGO format (tab-separated id+triples) (3.4Gb). This dump can be converted to different formats with the yago2converters (see below "Tools for Querying and Converting").

We also provide YAGO2 core in other formats, so that you do not have to convert it yourself:

YAGO2 core: Same as above in RDF(S) format (3.4Gb).
YAGO2 core: Same as above in the N3 format. This dump does not include reified meta-facts (0.5Gb).
YAGO2 core: Same as above in a format to use with the Jena TDB store, which supports SPARQL queries (1Gb).

Full

Full contains everything that is in core. Additionally, we include all the entities and facts from GeoNames - a complete import of the GeoNames data (from a dump of August 2010). It also contains textual and structural data from Wikipedia: all links+anchor texts between the YAGO2 entities, all Wikipedia category names (even those that are not used as a type in YAGO2), as well as the titles of references. In total, it contains nearly 10 million entities and more than 80 million facts.

YAGO2 full: in the default YAGO format (tab-separated id+triples) (8.7Gb).
YAGO2 full: Same as above in RDF(S) format (9.6Gb).

Tools for Querying and Converting

The converters run on the native form (tab-separated id+triples) of YAGO.
We provide a converter tool that can

export YAGO to RDFS, to Notation3, and to XML
load YAGO into a database (Postgres, MySQL or Oracle)
load YAGO into a Jena TDB store.

The tools also allow to query YAGO — either in the native query language or in SPARQL.
Download the converter tool (11Mb)

We also provide a tool to decode the native backslash encoding of entity names to UTF-8, which is also needed when you want to query for entities that are encoded in UTF-8.
Download the de-/encoding tool.

Linked Data

YAGO2 is also part of the linked data cloud, and are directly linked to by DBpedia. The links between YAGO2 classes and entities are also given in the following N3 files:

DBpedia-YAGO2 individual entity links (27Mb)
DBpedia-YAGO2 class links (4Mb)

Querying

There are multiple ways of querying YAGO2:

You can query YAGO2 directly online through the Web interface.
You can also use our Java interface to query YAGO2, see above under "Tools for Querying and Converting".
Use the SPARQL endpoint provided by OpenLink.

License

The YAGO2 Ontology is licensed under a Creative Commons Attribution 3.0 License by the Max-Planck Institute for Informatics.

YAGO Archive

You can still download the original YAGO(1) ontology here.