Or: How the New Year's release of f»dforward delivers more relevance through content matching but how users still determine the recommendations

Happy New Year everyone,

We recommend having a great 2009!

A New Year, A New Release

We celebrate the start of a new year with the launch of a new major release, the first sinds September 22nd.

Looking back a few months, we feel we have arrived onto a serious level with our technology. The first release covered basic collaborative filtering recommendation for blogs. We started with 6 blogs and followed a limited growth to some 20 blogs, all dutch. The new release introduces an improved recommendation method, multi-language support and a new level of scalability, enabling us to basically grant any request for participation for any international blog now.

So, what's new?

The latest additions were the flexible privacy options and the option to connect with and recommend blogs (for blog owners only!) that you admire, are friends with or is in any other way of interest to you.

To put it in one prase, the new year's release delivers the following to you:

More and more relevant, muli-langual recommendations in a more stable and scalable environment and first access to some backend dashboard functions for our bloggers.

How did we do that?

There are 5 major improvements in this release, being multi-language support, (semantic) content analysis, improved algorithms, a rebuild parser and a dashboard for bloggers.

Multi-language support

The widget now supports virtually any language. US English is fully implemented and Dutch of course, being our locale. Please send us your request for any language in which you would like to use f»dforward and for which blog of course. Use the (English) widget at the bottom of this page.

Content Analysis

Every new article that you write is first analysed for its contents. Through a few layers of information retrieval technologies we try to summarize your text in just a few words. More specific, these keywords are unique in the way they relate to other articles across the network. For example, check out what we make of this article in the widget at the bottom.

Though we still feel that the best keywords are tags provided by you bloggers, we try to come pretty close to that in an automated way. The keywords are used in our new recommendation algorithm.

New Algorithm for Recommendations

Normally, almost any article will have one or more content relations to other articles. We can check how much they are related based on their contents, but this is usually not what is actually read and found interesting by users. So, basically, we check if other people found content matches interesting too.

The concept of the algorithm is astonishingly simple. Relations between articles can be user based or content (keywords) based. We created a balanced combination for recommendations using both in just one algorithm. The algorithms looks for connections to other articles through users and keywords. One article can connect through more than one user or keyword to another, creating many matches. These matches are evaluated by there strength of connection and accordingly ranked. You get served three times the top 3 out of those matches: for local recommendations, in your network and in the f»dforward network).

So though we now support content matching for relevancy, your readers still determine the best recommendations. It is why we state: Content may be king, but users make a democracy!

You read that here first! Ok, no silly jokes now. On with parsing...

Rebuilding a Parser

Really not much to say about this. We rebuild it. The first parser could handle the small group of participating blogs very well. We of course hope for many more blogs to join the f»dforward network and therefore we needed a whole level of scalability and intelligence. Done that, moving on!

Bloggers In Control

We started on building a blogger´s dashboard. This dashboard already shows you:
» your blog's settings
» which articles we registered from your blog
» the main body text we use voor analysis
» the complete list of keywords we determined through this analysis
» a list of recommendations based on keywords
» a list of recommendations based on reader's interests
» a combined list of recommendations
» and the three small lists that will appear in the widget for your readers

These first steps should give you some insight in how we create recommendations. You will be able to proces your page again if something was changed or went wrong earliar. We have lots of stuff waiting to be deployed for you in this backend dashboard and the main purpose is providing you with insight in what is found interesting by readers on your blog, in your network and on the web in the general.

You can access it now. Open a widget and click on manage. You will find two links for managing your blog and managing your page. Click here to go directly to the manage page (bloggers only) in the widget on this page.

What is next?

Well, the things above are what we have been working on. In the next release you can expect that we pimp up the interfaces of the widgets, which are honestly lacking in some usability and flexibility, we think. In the coming releases you will also see:
- direct in-page quick recommendations (widget appears standard opened in page)
- easier participation for new blogs (no registration forms, plugins for blogsoftware)
- much improved flexibility and usability in design (another freq.asked request)
And, of course, quite interesting:
- the dashboard improvements (e.g. more insight in cross-network reading behaviour of your audience)

Join Now!

What, you have no free widget for your blog yet? We can't imagine, because we have it ready for you:

Request your widget immediately!

And don't forget to leave us a remark too. In this stage, your remarks really determine the way we are developing now. Thanks!