Skip to content

DBpedia Spotlight’s anniversary

February 15, 2012

Today we complete one year since we first released DBpedia Spotlight (v0.1). A small but committed team, 12 months later, and we’re still alive and kickin’! Our ambitious proposition to link text to *any* of the 3.5M things of 320 different classes in DBpedia has proven both challenging and a lot of fun. In our paper at I-SEMANTICS 2011 we showed that our simple and generic approach works quite reasonably, and that our focus on staying flexible pays off in the end. This perception has been echoed by other projects that have used DBpedia Spotlight throughout the year. We were excited to have counted 1.9M requests from 21/Nov/2011 to 08/Dez/2011, and managed to keep 98.64% up time since Aug/2011 up to now with 73 little availability accidents. 🙂 (By the way, thanks to Pingdom for tracking our availability for free.)

So much support from the community has made us both proud and agitated, as we would like to roll out more and more enhancements to support all of the fine people that have relied on DBpedia Spotlight in their work.

After our I-Semantics evaluation, we have participated at TAC-KBP 2011, which evaluates systems in their ability to link entities of type Person, Location and Organization to a reference Knowledge Base (KB). Their KB is much smaller than DBpedia, and the task is much more focused as there are only three target entity types. We have participated with DBpedia Spotlight as-is, and obtained surprisingly good results for a generic approach (around median), comparable to many old-timers in the field. Such a focused task allows for much more specialization, and we hope to participate again this year with a system specialized for that task.

We have also been working on developing new “spotting” techniques, that go beyond our lexicon-based approach. We have experimented with Named Entity Recognition, Keyphrase Extraction, among others. These results will be reported at LREC’2012 later this year. During that time we will also release some of the datasets that we created for DBpedia Spotlight, and give a talk on how they can be used to build or enhance NLP tools.

Next steps for DBpedia Spotlight include the release of internationalized versions, implementation of new disambiguation techniques and general enhancements (esp. performance of the ICF disambiguator).

Thanks to all of you that have supported, used or recommended our tool, and keep sending your comments our way! Special thanks to Jo Daiber, welcomed as new committer and Rohana Rajapakse and Iavor Ielev for their contributions in 2011.

Happy Birthday DBpedia Spotlight!!!

From → News, Uncategorized

Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: