Friday, October 2, 2009

Another Update: Now even more smarts in the client

We just released another new version of the Dispute Finder Firefox extension.

This one continues further down the path of pushing more smarts into the client. In the previous version we would only highlight a phrase if it was an instance of a known disputed claim. In the new version we use a simple textual entailment algorithm to highlight any phrase that is similar to a known paraphrase of a disputed claim.

The upside of this is that we now highlight much more stuff.

The downside is that we are now more likely to highlight stuff that isn't actually making the disputed claim - but we have plans to make it a lot smarter in the future, no whack down that false positive rate.

Once again, I'm afraid you are going to need to upgrade your extension.

Thursday, September 24, 2009

Major new version: Claim highlighting now phrase-based and done in the client

We've just done a major new upgrade to Dispute Finder.

The bad news: You'll have to upgrade your firefox extension.

The good news: It's much better and marks a lot more stuff.


Previously, the way that Dispute Finder worked was that the users used our search interface to mark snippets on the web that made a disputed claim. As you browsed the web, the Firefox extension would check whether the page you were looking at had been marked with a disputed claim, and let you know.

In the new version, the marking is instead done inside the Firefox extension. The Firefox extension downloads a list of phrases that are known to indicate disputed claims, and highlights them when it seems them on pages that you browse.

The advantage of this new approach is that, once a paraphrase of a disputed claim has been entered into our database, it will be highlighted everywhere it appears on the web - including on web pages that have only just been created - very important for news.

The current marking algorithm is rather dumb - it just looks for exact phrases. In the near future we plan to extend this to a more sophisticated natural language processing algorithm that takes account of the context in which phrases are being used and does not require that a phrase be used exactly.


Try it out. We'd be interested to hear what you think :-))


Friday, August 14, 2009

Hear me on NPR's "On the Media"

A couple of weeks ago I did a segment for NPR's On The Media talking about Dispute Finder. You can listen to it here:

http://www.onthemedia.org/transcripts/2009/07/31/02


On the Media is a really great radio show about what is going on in the media. Definitely worth being a regular listener :-)

Saturday, August 1, 2009

Updates to the web site

Hi Guys,

We just deployed a new version of the web site.

This version includes a lot of the changes that have been requested by users, and also fills in some features that probably should have been in the previous version.

Notable features include:
  • We've rejigged the look a bit. In particular, the tabs and search should be easier now and snippets look a bit prettier.
  • You can now more easily unmark snippets other people have marked. The intention is that this be used if you see that someone is spamming the system. This is combined with a background moderation system that lets us see when users are disagreeing over whether something is making a disputed claim.
  • You can /finally/ delete a claim that you created in error
  • This version includes an initial version of a machine learning algorithm that learns how to find snippets on the web that make a disputed claim, based on your examples. You may notice a robot face and a percentage next to some snippets in the snippet search interface - this tells you how confident the learning algorithm is that the snippet is making the claim.
There is also an accompanying new version of the Firefox addon. This will be available as soon as it gets through the mozilla review process. The main new feature with this update is that the addon now highlights only the text that makes a disputed claim, rather than highlighting entire paragraphs.

As always with a new version, it is likely that we have introduced some bugs. Please let us know if you find any by either commenting on this blog post, or clicking the "send us your feedback" link in the Dispute Finder web site.


Thursday, June 25, 2009

Press Coverage : Week 1

It's now a week after we launched Dispute Finder to the public and I'm pleased to say that we have been getting some pretty good press:

New Scientist did an article about Dispute Finder

Chris Dannen of Fast Company listed Dispute Finder as one of his tech highlights of the week

Rob Enderle of TechNewsWorld made Dispute Finder his product of the week

The Guardian (UK National Newspaper) said Dispute Finder was a "hit"

VentureBeat seemed to like our stuff

This is on top of past coverage from TG Daily, PC Magazine, and CNet, from when we gave a press preview of Think Link (the previous version) back in April.


We'll be maintaining a more detailed list of press at on the Confrontational Computing Press Coverage Page

Welcome to the Dispute Finder Blog

Hello and welcome to the Dispute Finder Blog.

This is a space where we will talk about what is going on with dispute finder, announce new features, and solicit feedback from you, the users.

Dispute finder launched to the public only last week, and we have already been getting some good press and good content being added by our users. Rest assured that what you see now is only the beginning. We have lots of fun stuff planned which will hopefully be arriving in the coming months.