Linking Open Conference Tweets

From Semantic Portal Wiki

A Linked Data-a-thon mashup by Joshua Shinavier, Li Ding, and Zhenning Shangguan. Its twitter hashtag is #loct

Top ISWC 2009 twitterers (by number of ISWC-related tweets)
Top ISWC 2009 events (by number of twitterers tweeting)	Top ISWC 2009 events (by number of tweets)
Top ISWC 2009 topics (by number of twitterers tweeting)	Top ISWC 2009 topics (by number of tweets)

What is this?

This demo mashes up a couple of semantic microblogging services with the Semantic Web Dog Food Corpus, Twitter's search API and Google's visualization API to produce live views of hashtag usage in tweets about this conference. The key component of the mashup is user-driven Linked Data extracted from microblog posts. Our starting point for the visualizations is a single resource: http://data.semanticweb.org/conference/iswc/2009.

From there, we follow semantic links to sub-events of the conference with a combination of OWL inference and SPARQL queries. We also follow user-contributed semantic links parsed in near-real-time from the Twitter stream. We look specifically for posts using a select few semantic nanoformats in tweets like this one:

   Watching Danh Le-Phuoc's talk on mashing up sensor network data w/ Semantic Web at #ssn2009 (see http://bit.ly/303WmQ). #realtimesemanticweb

and this one:

   final wrap-up with @joshsh on #ldthon (part of #iswc2009).    

but right now, we're just using owl:sameAs links from tweets like this:

   Linking #sdow2009 (same as http://data.semanticweb.org/workshop/sdow/2009 ) into SW Dog Food.

We currently listen to a couple dozen individuals for structured tweets: please email us if you'd like to be added to the list! A service called TwitLogic aggregates tweet-embedded data and makes it available as Linked Data, which allows us to discover relationships among hashtags which are stronger than simple co-occurrence.

Visualizing real-time, user-driven semantic data (in nine kludgy steps)

1. Grab the ISWC 2009 Semantic Web Dog Food data dump
2. Combine with dynamic, user-contributed TwitLogic data harvested from the Twitter stream
3. Perform symmetric and transitive closure on owl:sameAs links to simplify our SPARQL queries
4. Select interesting hashtags with SPARQL, making use of user-contributed semantic links via TwitLogic
5. Query the Twitter's search API for tweets containing the hashtags
6. RDFize the tweets using an existing twitter4rdf tool and store the results in flat files (no this won't scale. No, we don't care!)
7. Generate SPARQL queries for the combined data set
8. Execute SPARQL queries and convert the results to JSON for the Google visualizations
9. Watch the statistics change (every hour or so) as we all tweet about the conference. A minority of "semantic" tweets add value for everyone.

Still to do:

1. Allow users to navigate through the visualizations to the linked data for people, conference events, hashtag-based resources, and anything you can name with a URI

Implementation

Here's the source code. It's a mess.

Status (Twit #loct)