Tuesday, January 15, 2008

Tinfinger beta starts today

Today marks the beta launch of Tinfinger, after two years of off and on development by myself and Tai Tran. It's a very exciting day, if a bit sleepy because we timed it for 9am Californian time, so it's 4am local time here in Geelong! The launch press release:


Tinfinger changes the rules of Web creation

Human omnibus launches with user-debt strategy breaking new ground between Wikipedia and Squidoo/Mahalo/Knol

(Geelong, Australia, 15 January 2008) -- Fans of famous people will have a new place to share their fandom with today's beta launch of Tinfinger.com, a human omnibus. The site combines user-authored encyclopaedic profile pages of famous people with a news and blog search engine based around mentions of those peoples' names, which are aggregated into frequently-updated front pages for 650 categories. Tinfinger is intended by its two Australian co-founders, Paul Montgomery and Tai Tran, to become the primary resource for information about famous people on the Web.

"Tinfinger will be to the Who's Who what Wikipedia was to the Encyclopaedia Britannica," said Montgomery. "The Web is ready to move beyond the hyperlink as the only way to cluster and rank Web content, and we're going to try using names instead."

Tinfinger has not taken funding, which led Paul and Tai to devise a new business model: going into debt to its users. Contributions to Tinfinger will be paid for not with cash but with Google AdSense impressions (AIs), so that Google pays Tinfinger users for ads that Google puts on the Tinfinger site. It is expected that the rate of page production and AI payments will outstrip ad inventory at first as traffic to the site builds gradually, so that Tinfinger will start with a debt owed to its users, payable out of its future page views. More on the AI system below.

"This is an innovative business model built on necessity. We hope those who choose to participate develop a sense of ownership not only over their own contributions, but over Tinfinger as a partner in an ongoing contract," said Montgomery. "There are plenty of people out there who would like to meet other fans of their favourite people, and we hope to create a way for them to share their passions."


1. A database of famous people - famous meaning that they are mentioned in news or blogs related to a newsworthy issue - which classifies people both via a top-down category structure and a flat tag structure. Tags on Tinfinger will be expressible as RDF triples (subject-predicate-object, as opposed to subject-object). At beta launch time, existing people and their tags are not editable by users, but registered users can submit new people.

2. Profile pages about famous people, featuring articles and pictures submitted by users using WYSIWYG content authoring software. A collaboratively-authored profile article of around 150 words for each person, similar in function to a stub profile on Wikipedia, will be released by Tinfinger under the Creative Commons license. Each profile page can also include many other types of copyrighted single-author articles: biography, review, interview, encounter, comparison, praise, criticism, etc.

3. Clustered news aggregation with "front pages" for 650 categories, with a design familiar to readers of Google News, Topix or Techmeme. Tinfinger operates its own news and blog search engines, and snippets from these sources are collated into clusters based on mentions of each famous person's name. Tinfinger does not use links or semantic connections to cluster; just names, using a publicly available algorithm called tinscore. Higher-level category news pages include people from lower level categories, so that for instance the Africa category contains stories about people from all African countries, and the Internet category contains stories about people from Search Engines, Web 2.0, Web Advertising, Voice Over IP, Broadband and so on. Users can submit new sites, and the list of sites indexed for each category is available as an OPML reading list. The news and blog searches (as well as articles and pictures) each have their own RSS feeds and also can be published to other sites using a widget.

4. Social networking software from PeopleAggregator to enable user interaction and feedback. There are Tinfinger-controlled groups for each category, and users can create their own groups.


In the AI system, users will be rewarded for writing articles by their AdSense publisher IDs being put on Tinfinger pages. This is not a new thing by itself, but existing systems involve giving users a percentage of page impressions next to their articles. Tinfinger will debit the user's account with a fixed number of AIs for each article, which will then be gradually paid out of traffic on Tinfinger pages - not next to their articles, but from general site ad inventory. (The category headline and profile pages will be reserved for Tinfinger.) The starting rate for articles will be 10,000 AI. It is unknown what CPM rates that ads on Tinfinger pages will attract, but Tinfinger is aiming for at least US$1 CPM, implying a base payment per article of US$10. The AI figure each article actually earns will be highly changeable based on various quality and editorial factors, which are determined by Tinfinger based on published rules, and can also be boosted if Tinfinger places temporary "bounties" on articles about particular types of people. This approach will likely lead to a significant AI debt which Tinfinger will owe for many months. That will be part of our partnership with users.


- The co-founders and only employees are Paul Montgomery, a former technology journalist, and Tai Tran, a former corporate programmer. Both live in the Australian city of Geelong, which is 100km southwest of Melbourne. Paul blogs at http://tinfinger.blogspot.com and has blogged quietly about many Tinfinger features already.
- Development of Tinfinger began more than two years ago.
- Tinfinger has taken no funding, and is not currently in the market for funding. It will effectively be funded through debt owed to its users.
- To discourage spam, the social network portion and external links from profile pages will be labelled with nofollow.
- Tinfinger will consume OpenID logins.
- Database figures at launch: 404,000 people, 395,000 snippets, 2,700 pictures, 612,000 tags, 820 sites... 10 articles. Most people records were adapted from dbpedia and IMDb; most tags were from dbpedia. The people database is lumpy, with some categories containing very few people as yet.
- Tinfinger's mascot is a robot called Ned, who bears a worrying resemblance to Sidney Nolan paintings of the Australian bushranger Ned Kelly.
- The Tinfinger news algorithm, called tinscore, is detailed at http://tinfinger.blogspot.com/2005/12/tinscore-and-other-ways-to-clone.html


What to do at Tinfinger


Links to media coverage
TechCrunch: Tinfinger: A User Generated Who’s Who


Post a Comment

Links to this post:

Create a Link

<< Home