Comments for SemanticProxy: Jump-Starting the Semantic Web

SemanticProxy: Jump-Starting the Semantic Web

2008-09-23T15:19:34Z

While it has great potential, the Semantic Web has failed to live up to its promises so far. Part of the problem, as Thomson Reuters sees it, is that developers will not add a lot of semantic features to their products until publishers start publishing more semantic data. Reuters' OpenCalais represents one way around this problem. But starting today, Reuters' newest project SemanticProxy will give developers an easier way to extract semantic data from any web site.

]]> Even though SemanticProxy is geared towards developers, Reuters has created a demo site that you can try out on the web by just copying and pasting the URL of any web page into a simple form. We tested it with articles on CNN, Wikipedia, and a number of blogs, and it always returned a highly relevant set of results (as long as the page was not excessively long). The service is optimized for performance on 30 of the world's largest news sites, but it also works just as well for other sites.

For a news story, for example, SemanticProxy will identify politicians, cities, countries, etc. that are mentioned in the article. Once parsed, the service returns the semantic metadata of the page in three possible formats: RDF, MicroFormats, or standard HTML.

As the name implies, SemanticProxy acts as a proxy and aggressively caches all its data, which should make it easy for a developer to scale a project that relies on this service.

Catalyst

SemanticProxy is part of Reuters' attempt to jump-start the semantic web. As Tom Tague, the leader of the Calais initiative at Reuters, points out, SemanticProxy can hopefully act as a catalyst and get more developers to look at semantic data, which, in return, will give more developers a reason to publish this data themselves.

Disclosure: Calais is a RWW sponsor

Comment from latifah on 2008-09-23

2008-09-23T20:44:03Z

Although the potential of such a service is really great, it is still not accurate enough to build serious applications on top of it.
A few examples from a jazz site processed by calais:

Person Political
person: Latifah
position: Queen

Technology
name: Gershwin

Person
name: Be Good

Doing accurate semantic analysis seems like a really hard problem. Although the "semantic web" community claims that the technology is ready and we are only waiting for the killer application, the truth is that we are waiting for the killer technology..

Comment from Josh Kirschner on 2008-09-23

2008-09-24T05:30:24Z

As latifah clearly demonstrated, any semantic technology must be measured first and foremost on the accuracy of its results. This is accomplished both through the technology and by the team of linguistics experts who sit behind that technology to build and maintain the underlying taxonomy.

Companies such as mine, Inform Technologies, have been providing semantic solutions to the web community for years (and our technology knows that Queen Latifah is not a member of the royal family). Serving our clients with the best semantic technology is our sole mission, not selling more news feeds or developing proprietary portals that compete with our clients. With dozens of Inform implementations at leading media sites, across a wide range of industries, this isn't just a vision - it's today's reality.

Comment from Zach Beauvais on 2008-09-24

2008-09-24T10:29:36Z

I know this service is designed to be used "programmatically" (i.e. used by software rather than executed by hand by a person), it does show an incredible potential to a general user.

I did this for my own site, and the results were mixed, as reflects a beta trial. Also, they don't YET produce dereferencable URI's for the data (though, as Paul Miller over on ZDNet pointed out: it's on it's way). I think as soon as a user can fill in the simple web form, and receive a working RDF graph of their site, it will be incredibly exciting!

Comment from latifah on 2008-09-24

2008-09-24T10:41:11Z

Josh, if this is the case, it would be great if Inform Technologies will make a web service similar to Calais

Comment from Zach Beauvais on 2008-09-24

2008-09-24T12:37:00Z

I know this service is designed to be used "programmatically" (i.e. used by software rather than executed by hand by a person), it does show an incredible potential to a general user.

Comment from Zach Beauvais on 2008-09-24

2008-09-24T12:38:17Z

I know this service is designed to be used "programmatically" (i.e. used by software rather than executed by hand by a person), it does show an incredible potential to a general user.

Comment from Zach Beauvais on 2008-09-24

2008-09-24T12:40:28Z

Sorry, I'm not entirely how that posted three times... the FriendFeed was sending my comments to someone elses' feed, however, so might be a good idea to reset it?

Strange...

Comment from Atul Kedar on 2008-09-25

2008-09-25T14:58:02Z

This is a great stop-gap tool, until the content entry tools start managing content in ways that can easily be transformed as rdf/rdfa/microformats during the presentation cycle. Since the content authors are in the best position to assist in markup, they should make the initial attempt. This markup need not be complete in any way, but if some of the easy ones can be handled - user info as foaf, location as geo, events as hCal, tools like senaticproxy can more accurately fill in the missing semantic tags.

The troubling aspect of semanticproxy is the intellectual property consideration of semantic info added by semanticproxy, do authors want to give up their rights to metatags used to semanticproxy. Since the service is free for now, there is a legitimate financial need for Semanticproxy to be able to use the resulting information in some way and monetize the operation indirectly. My own opinion on this is unclear and would like to discuss the implications of using this service by content producers.

Comment from Greg Boutin on 2008-09-28

2008-09-28T14:14:42Z

I have commented on this, both praising and raising questions, on my blog at How Semantic is Linked Data? OpenCalais Launches Semantic Proxy

One I haven't heard back in particular, is what type of RDF attributes does Semantic Proxy create beyond the verb "To Be"? RDF is meant to include a vast range of linking possibilities, but I don't see how OpenCalais (the platform behind Sem Proxy) could add qualified relationships.

So far this strikes me as a very nice application of the Linked Data concepts, less so a semantically intelligent one.

Any insight welcome.

Comment from Tom Tague on 2008-09-28

2008-09-28T17:08:18Z

Tom Tague from Calais here.

Thanks for all the comments - it means people are trying things out and thinking about the issues - which is what we want.

Clearly SemanticProxy is nowhere close to the end game for semantically enabled sites. I'd encourage people to take it for what it is - a lightweight mechanism for getting a good semantic overview of a site - and one which will continue to improve over time.

A couple of points - Atul - we don't ask anyone to "give up their rights" to the metadata. We need to retain copies of the metadata if we're to have de-referenceable URI's - no way around it. You retain ownership - we just have some rights to use it.

Greg - you're right. At this time we're primarily identifying entities, facts and events with a "is a" type of identifier. We can't hope to extract all of the semantics incorporated in a page - and frankly don't want to. We'd rather be high performance plumbing and let our users go off and do very cool things with what we've identified and the linked data that (will - 4Q - We Promise) points to.

Latifah - oops. That stuff happens when you're running fast. I'd encourage anyone who finds these sorts of things to drop us a note at errors@opencalais.com and we'll get it on the plan to get fixed.

No comment on the advertisement except to say feel free to jump in and make your tool public. You'll be amazed at the productive feedback you'll receive.

Regards,

Tom