Bernhard Seefeld's Blog

Google += Endoxon;

I just acquired a few thousand smart coworkers :-)

Links, Back-Button and other Web Principles in AJAX

Here is a presentation (in german) on Maintaining Web principles in AJAX that I'm going to give in about one hour at the EuroAjax 2006 conference. It is a co-presentation with J�rg Stuker who covers SEO aspects of AJAX.

I cover deeplinking into AJAX applications, supporting the back-button both inside an AJAX application and back-into-page. The described techniques should cover IE, Firefox and (recent) Safari.

See also J�rg's Post, who also offers a downloadable version of the talk.

Tags: backbutton, permalinks, ajax, webarch, talk

goyellow.de/map launched

Finally, I can announce, that the first customer from my new project launched this week! Working together with Endoxon AG, the new online map for germany at goyellow.de/map is now online. Please go and have a look!

You will find a lot of similarities to map.search.ch, with a couple refinements here and there. It is a full AJAX map for Germany with hybrid and vector views. You can overlay a lot of information from GoYellow, a internet yellow pages site for germany with direct door-to-door public transport directions available to every such point. Additionally, we have global data for the first few zoomlevels, so you can get a decent view of the rest of the world.

Of course, you have nice URIs like goyellow.de/map/berlin or goyellow.de/map/muenchen/hildegardstr-1, so I wish happy linking!

We put a lot of work into the visuals, especially the image quality and the general UI behaviour. Please have a close look! Note the clarity of the colors and the labels. Or drag the zoomslider.

The project will move forward adding more countries in this way. And soon, we will have some new concepts waiting for you. Oh, and we will get an official name..

More soon!

Update: My coworker Gregor shows how you can link to all kinds of information layers. And it seems like we already got first sort-of mashup.

Tags: mapping, germany, ajax, goyellow

Classifieds from the Edges

After a few posts on aggregating "Web 1.0"-style classifieds, I finally have something to blog on a possible next generation of classifieds:

At last, there is some news on edgio, Mike Arrington's new venture. As expected from the little we knew before it is primarily an aggregator of classifieds on blogs. Somehow, I wondered why that took so long, since the idea has been on the table for quite a while, but on the other hand there is not much selling-via-blogs going on yet anyway. I hope that edgeio and similiar sites will change that.

From the description it sounds like a solid and complete implementation. They ask sellers to use the 'listing' tag to draw attention, and although the article claims that they will "constantly crawl millions of blogs", my guess is that they subscribe the technorati feed for that tag. In this way, they re-use all the anti-splogs mechanism technorati implements. That would be a very simple and neat model to aggregate things, one that could be copied for many different things, too. In fact many geoblogging sites rely on one way or the other on such a mechanism. And of course, for events it is commonplace now to agree on a tag, to aggregate all related material. Edgeio would - if I guessed right - take this one further, by using the tag only as a means to find a URL and then going further and actually download and analyze the page.

Since structured blogging and microformats seem like perfect matches for edgeio. I wonder wether they even demand the implementation of one of these models for posters in order to get some structure. But if they do, then they should really mine all blogs and use the listing tag only as a speedup mechanism (i.e. being the Web 2.0 equivalent of 'Submit URL' at the search engines).

The article mentions an interesting usage of trackback. It sounds like they will automatically trackback every post they find and include. While that sounds pretty intriguing, I fear it can quickly become very problematic. Essentially, the whole process is automatic mass-trackbacking of all people using the listing tag. Sure, there is value in this trackback, but where do you draw the line and who decides what useful is? And what if there will be 20 more aggregators? I think that every form of trackback that doesn't result from a human choice on the trackbacking site (usually hand-picking a link to the trackbacked site) borders to spam, especially if high volumes are in the game. Well, as the other of the article says, he is not sure that he got every detail right, and the people involved in the project are definitely deep thinkers and well aware of the good and bad potential of these techniques, so I stop speculation and have another look after the launch.

By the way, here in Switzerland, Kaywa is doing something in a similar market with ichiba. There aren't many details out yet, but from the little I heard they seem to take a slightly different approach, that might sit a little closer to the way classifieds work for users today.

Tags: classifieds, disruption, aggregators, listing :-)

The Perpetual Beta: Is Webdevelopment Stuck to Small Incremental Updates?

The term "perpetual beta" sums up an important feature of web based applications: You can roll out updates as often as you want and all users immediately profit from it.

But once a site attracts a non-trivial userbase, this might also be its achilles heel: It can be hard and risky to impose major updates on your regular users. Sure, you can add a major set of new features, but that leads down to the path of complicated sites. But if you want to progress by changing how a current feature works, then you can get yourself in trouble. To limit risk, only small changes are considered. Which is smart for many other reasons, too, of course.

But what if the web around you progresses and a major shift of course is in due? What if your userbase is nicely in the early and late majority. With shrink-wrapped software you could jump on the next curve with the next release, sell it to the early adopters and give the rest some time to get comfortable.

You can't really do that with a website. Sure, you could add a special beta section an run that in parallel, but only real innovators go there and you'd have to keep it separated quite a long time, would loose even more power to keep the majority of users and if your site has social aspects cut out the the network effects.

So, you have a choise: Stay at the top of major shifts and risk alienating the most faithful users or eventually get caught up from behind by a competitor starting from scratch or coming from another field.

It was impossible for Yahoo to radically cut down their Homepage when the early adopters rallied around Google. Even now, their AJAXified Mail (guess why it is taking so long since the oddpost acquisition!) will keep all major principle in place. Gmail was able scrap folders and replaces them with tags.

If you understand german, you can watch the dilemma in the comments of the immo.search.ch launch. While it was in beta, there were 100+ mostly positive comments, in fact extremly positive. Once it was officially launched, that is, it replaced the old immo.search.ch - a pretty standard form -> resultspage -> detailpage type site - by a new ajax-centric site, old users took the general tone in a very negative direction. They want the old system back. They obviously copy arguments, which you can easily spot: One user mentions an actually quite obscure bug that appears when you search for a house to rent (and not a house to buy or a flat to rent, the majority cases) and that was also present in the previous version, as it is bogus data not code. But it serves them as an argument, so they use it. Fascinating to watch the dynamic in the comments.

Now search.ch is bold enough to hold to the new concept. They believe that it is a genuinly new appraoch that will ultimately beat the older system. But the real estate search was never a really important for search.ch, so they can take the risk. The commenters in the blog are probably a vocal minority and I would guess that a reasonable part of the current users will stay and even spread the word. But even if the complete userbase would have to be replaced with this step, the payoff vs risk made sense.

But not so for the other big real estate sites. For them, the current userbase is one of the core assets. And every step that might piss off a good part of it is a bet-the-company step. At the same time, they will surely loose users to immo.search.ch and other companies that are freeer do innovate in bolder steps. The best would probably be to time the transition to a new paradigma to a sweet spot where they can still get back a couple of users and really keep some so far loyal users that at least saw the new style and probably get used to it and the users they would loose are on balance. But they will come out hurt one way or the other.

Shrink-wrapped software with its granular update cycles and a self-adopting userbase don't know any such risks. Only to miss out an important change of direction, i.e. to not innovate enough.

So what could a site do, that built its success over the years with a model that is bound to be replaced over the next couple years? Are there any good examples where a site mastered a difficult transition in a fine way? What could be done by clever communication in advance and during the transition?

(All of that assumes that you know what you do and don't innovate in the wrong direction. That of course is whole different topic to analyze. For small steps in the wrong direction that perpetual beta is actually more forgiving - if you see the signs early enough. And you see them correctly, which is not easy: Just go and read the comments in the search.ch blog entry linked above)

Tags: innovation disruption ajax

No market for Mobile Location Based Communication Systems?

After a data-rich and fascinating talk about the specialization of communication channels at LIFT yesterday someone asked if the introduction of position-information of your communication partner would alter the communication patterns. Stefana's suggested that there probably would be not much change in the pattern, since the 4-6 people that you do the bulk of synchronous and nearly synchronous communication with, you know well enough to know where they most probably are at any time anyway. Taking this further, these would of course be the same set of people that would probably give you the trust rating to expose their position. Thus, the question: Is there any market for all these mobile location based communication systems?

Not in the obvious implementation case. Maybe either by restricting types of locations, like Plazes does (which maps places I'm online, thus probably working and almost never my spare time) or by coming up with a system that is anonymous enough, but yet useful and not annoying (permission spam) so that you could extent the service to a much wider circle.

On the other hand, IM-style presence indicators (from "away" and "busy" to "phone", "meeting", "in the zone") potentially fulfill a much more important note in this context. Maybe phones start listening to things like how many voices (and at what volume?) there are in the room, how fast the typing-sound is, etc. They could at least switch to silent mode in this case? Hmm, how reliable could you detect that the owner is sitting in a movie or a talk? And why is there no "silence please" beacon installed in every cinema that mobile phones could pay attention to?

Tags: lift06 lbs socialsoftware mobile

Network Latency as Boringness Indicator at Conferences

I'm currently at LIFT. And I have a theory:

The interestingness of a talk is inversely proportional to the latency on the conference wifi.

(This post will take particularly long to get to the server)

Update: 1. Most of the conference is actually quite interesting 2. Anina is coming up, I expect low latencies this afternoon. 3. Dang, do I sound geeky today.

Tags: lift06

New AJAX Swiss Real Estate Search

search.ch just launched the new beta real estate search (immo.beta.search.ch). It is one of the last projects that I started when I was still CTO at search.ch and it's great to finally see it online. See the post on the company blog (in german) for more details.

It aggregates objects from about 20 real estate sites in Switzerland, constantly adding more. It even includes sites that offer rooms for students.

It's AJAX all over the place. With neat details like to incrementally loading result list on the left, the small bar charts above the sliders (so that you know wether you have to allow higher prices or smaller flats, or just get a general feeling for an area), updated in real time.

You will see all available objects in the map, and the ones visible in the list at the left are highlighted (try scrolling and see how they change). Actually just missed objects are also still on the map. One of the neat feature the team added after I left is the behavior there: The non-matching criteria is highlighted red and when you click it the according slider is adjusted to include this object.

Another nice part is the list management. Looking for the next flat or house is a process that goes on for a longer time, so you'll want to keep track of what you liked, tossed out again and so on. You can mark objects and you can keep notes. Best of all, when you return you get the exact same query that you had when you left last time and all the new objects since your last (and the one before) are separated by a marker.

The search.ch team did an excellent job, I'm proud of them! Congrats!

Update: A small translation table until the site is translated:
Wohnung, Einfamilienhaus, Ferien, Mehrfamilienhaus, B�ro, Garage, Grundst�ck: Appartment, Single Family Home, Cottage, Multi Family Home, Office, Parking lot, land
mieten, kaufen: rent, buy
Preis, Zimmer, Fl�che, Ort: Price, Rooms, Area, "Where".

The Classified Market through the Market Lens

Andreas G�ldi has a brilliant analysis on the classifieds market as follow-up his earlier and my post (thanks for the kind words!).

He nicely structures the discussion: He observes that we are talking about a market, and as such a system that works better when liquidity increases and transaction costs decreases.

Many of the suggestions in my last post can be catalogued under the transaction costs aspect, which are mainly outside the actual cost to publish a classified, just as Andreas notes. Thinking about it, lowering could grow the classifieds market manifold; or maybe this already happened, just look what is sold on ebay that nobody would put in the classifieds section of a newspaper.

Then Andreas splits liquidity in two factors: Presence and positioning. Presence, i.e. the cost of just being in the market, is quickly driven to zero by the aggregators and things like Google Base (Forget about usability gripes, the important point of Google Base is really to lower the entry barrier to put stuff on the web, something that will drive Google's bottom line - along with ~~all~~most other's bottom lines - in a much more significant way than anything to do with competing with craigslist. [Update: Or so I thought... Apparently Google is going the wrong way here]). Full agreement here.

But on the positioning, I'm not so sure, at least in this case. His point is that one could essentially sell ranking, exchanging money for a better position in a market. But for many important categories I don't have a problem specifying my query precise enough, that I can browse through all matching ads, thus ranking is not important. It would be easy to look at all classifieds for flats in Zurich in a given size and price range, and in fact many people do that and currently go to all the sites and read all the papers to not miss any.

(Yes, ranges are actually not that useful tools, but users usually give extra room to include ads just outside their range to double check. In fact we tried relevance ranking for real estate for a couple of years in immo.search.ch - you can just enter a target price and size and the engine will rank ads by proximity to this values, giving cheaper/bigger a slight preference - and it works quite well but in the end never got the traction with the users, thus in summary I think the idea was rejected by the market. And in fact the next version of immo.search.ch will swap these features out for something else...)

Back to positioning: Of course, one could sell the appearance of non-matching (maybe by a short margin) ads. Of course strictly separated from actual search results, please. But then again, a good site would be able to generate good suggestions here anyway and besides, in the classifieds case this looks like a rather small opportunity.

Thus, positioning so far makes sense where

the number of results is huge (web search engine, where the top 10 are critically more important than all the others) - i.e. almost equivalent to where relevance ranking is important
search quality is generally bad and almost more a means to stumble around and get ideas (e.g. ebay, where you can pay to increase attention to your products)
transaction costs for the seeker are high (thus artificially pushing the barrier for "too many results"-rule much lower), e.g. on ebay "which of the zillions ipod offers should I bid in?" or when the next step would be applying for a job. This only applies where I can't distinguish offers otherwise, either because they are nearly identical (ebay) or because of lack of further information (jobs).

"Too many results" is subjective and while some wouldn't mind skimming real estate ads half an hour per day, for others this is too much. But since the cost of a suboptimal decision outweigh time saving by just taking the first offer, these persons rather create an own market for assisting them (Andreas mentions this, too). And tools and methods for increasing skimming speed and accuracy will be what differentiates future aggregator sites.

Thus, I think positioning isn't where the profits will move in the case of classifieds. If, then only for the jobs market, which exhibits its own very complicated dynamics and very high transaction costs; my gut feeling is that this part will take different directions than all other categories in a big way anyway. Sounds about right, since that is very the aggregators will be and there (unlike for general search like Google) entry costs will be low and thus profits will move somewhere else in the value chain. Which doesn't mean that the operational profit for these outlets won't be handsome, but they most probably won't take the billions that are left on the table by slow moving newspapers (sorry guys).

Will the money be in lowering transaction costs? And if transaction costs go down and the market grows, who will profit? The sellers, the buyers (or the new class of seller-and-buyer), the new middlemen?

On Website Language Selection and URI Construction

J�rg blogs about language selection on web pages over at the namics blog (in german). He wrote up his guidelines, most of which are quite sensible, but also read the comments for a good discussion.

As a contribution to the discussion, I'd like to explain how we did this at search.ch: In general, very similarly to his guidelines, i.e. the language selection is in the upper right corner and the labels are text and in the destination language (a few pages still have the selection in the history/navigation bar at the left, but the upper right corner is really where it should be).

But the interesting details are within the initial detection and the URIs. An URI could end in help.html, but also in help.en.html; exactly how Apache content negotiation suggests, i.e. the first one would trigger automatic language detection, the second one is fixed on english. The important difference is in how the links in these pages look. We took great care (and credit for this goes to Urban M�ller) to keep the links on those documents in the same style. That is, the help.html document would come in the user's language (e.g. german) but wouldn't contain language links (e.g. .de.html). Thus, the url would also reflect the information the user put into the system: The language was only set if it was actually overrided by the user. Since the language is usually guessed right, this keeps nice and tidy URIs (directory indices go from /index.de.html to /!) for most users, and even more importantly doesn't leave to much information in the URI when it is used in a link or an email! Unless the user explicitly set the language, automatic language detection will still work when the user comes from an external link.

Another detail was, that we override an automatic detection of english with german. This might be a special situation that suits only for search.ch, but let's look at the reasons. Tests (comparing the language of queries with the detected language) showed, that english-instead-of-german was the most common mistake in detecting. Quite obvious actually, since still many people prefer to install their computers in english (I do) and then don't change the browser settings. And german is the most common language in Switzerland. And this wrong detection is quite a loss since many parts of the site (e.g. news) are not available in english. On the other hand, by the nature of the site search.ch an english speaking user who doesn't understand german would still get quite far in the wrong language. Thus, the false positives for english were worse than the false negatives are now after the overriding.

Of course, both differences are small details in the big picture, but for a fraction of the users getting those right matters a lot. As does hiding the details from all the others.

What does Licensing a Specification under Creative Commons Actually Mean?

Like RSS and other formats Microsoft today released their draft specification for Simple Sharing Extensions for RSS and OPML under a Create Commons license. I will dig into the technical details later, but it is clear that should this gain some traction, it will be a quite significant contribution to the web (An earlier, similar draft specification unfortunately didn't get much attention).

But for now, I'm interested in a much more basic question: What implications on implementors does licensing of the specification under CC have? I mean, people are making quite a fuss of Microsoft using that license, but what is the actual benefit for me as implementor? Sure, there are certain things I'm allowed to do with the document (i.e. the form) per se, but what can I do with SSE (i.e. the information described in the document) that I can't do with something that is specified under a "All Rights Reserved" document? E.g., I would allowed to describe the protocol in my words anyway and I could choose any license for that version. It wouldn't be the official version, but neither would any CC-allowed derivate be. There seems to be limitation (i.e. no attribution or share-alike necessary) on the license of the code that implements SSE (thankfully). So far, I don't see the special case for specifications.

I can see that it is extraordinary, that Microsoft uses an essentially GPL-like license (the share-alike is what Microsoft calls a viral license), but they use it in a place where there seem to be no relevant practical consequences of doing so. There is nothing really "non-commercial" about it, as some claimed. The copyright and the license cover the form of the specification, not what is specified. And licensing the text under CC also has no apparent consequences on the patentability of the underlying ideas. So essentially I can sing the specification or print it on a t-shirt and I can modify it before doing so. What else? Why is everyone so excited about it? Even Tim Bray seems to take the CC-license as guarantee of it being legally unencumbered. And Tim obviously knows a lot about format specs. So, do I miss something?

Can someone help me out here?

Update: Philip Jacob asks the same question. No clear answer yet, but it seems that the CC-part in this context is just good PR and not much more. So then, the question remains what the lincesing-unrelated part about patents and Microsoft means. I find the formulation ("... agrees to offer a royalty-free patent license on reasonable and non-discriminatory terms and conditions ...") pretty vague, although it at least sounds as meant in good spirit. Others have a less-optimistic interpretation here.

The Classified Market and the Aggregators

Abstract: Off- and online classifieds business models are based on selling access to a scarce resource, readers. While classifieds aggregators (like simplyhired, Oodle and others) do deliver free traffic to those sites, it isn't nearly as valuable, as it the value-generating property (scarcity) is stripped off: Those users would find the ad through any different site, too. However, blocking aggregators would be short-sighted, since even a market leader would be bound to only loose market share over time. The era of selling vanilla ads to scarce audiences will be over sooner than many wish, the base cost of publishing ads vanish. This leaves a huge gap in monetization and the key strategy to capture this money would be to base the value proposition on something one level up the stack of hosting ads. The article quickly discusses a few such ideas. Further, it seems like providing a strong API, i.e. helping the aggregators, would be a smart move for these sites, as this would spur necessary innovation and most importantly keep power-concentration at the aggregator sites in check.

Recently, a number of startups tackle the aggregation and search of classifieds over multiple sites. Switzerland will see its share coming, too, as in the beginning of this year a landmark court decision cleared the cloud of uncertainty behind this model -- interestingly one of the suers (tamedia) was the first larger company to launch such a site, piazza.ch.

Naturally, the classical classifieds sites are wary on the situation. The aggregators argument is, that the classified sites are profiting themselves, because they get traffic for free. While this is certainly true, the argument is also a bit disingenuous, as the aggregators move themselves into an interesting position, and this with relatively little effort. An interesting position, because they're the ones who can capture most of the user's click flow and slowly gaining market position to sell advertising or even premium positions in the result list. In essence, they say "we give you free traffic to your classifieds but want to show the search results page in exchange". And Google demonstrated how financially interesting that page is.

That said, the current classified sites have not been very innovative with their results page. The business model is just an almost 1:1 copy of the offline model, money is being exchanged for exposure to an exclusive (scarce) audience. The offline model has the audience' scarcity built in. The first wave of the remaking of the classifieds market was that the audience moved online; the roles of offline publications replaced by online destination sites. Newspapers lost market share by clinching too long to their old, profitable offline business. But all in all, not much innovation has happened in this market, surely not much was driven by the fact that we're online (except for the convenience to reach the site, of course).

The aggregators are removing the scarcity, the traditional reason for paying for the ad. In principle you can put your ad on the cheapest site that is being crawled, that is the price will quickly plummet to zero.

A rather ignorant reaction of a classified site could be to technically block all kind of spiders (Craigslist did, but apparently more for technical reasons). While that would slow the proliferation of aggregators, smaller sites would still be happy to be crawled. At the very least when the big portals start in this business. A market leader site would retain their scarcity for the moment, but would be bound to only loose and loose market share over time. This model is not sustainable.

As a classifieds site, a simple answer would be to jump on the bandwagon. Use your brand recognition to have a head start in the search business, and show your own customers at the top of the result list - the most obvious monetarization scheme. In fact, the launch of piazza.ch here points in exactly to that direction and several job sites have been doing this forever. The elephant in the room here being the question wether a half-hearted approach, especially regarding the ranking, will be strong enough against new sites, that focus on being an aggregator. Remember, the barrier of entry for starting in that market is low and getting lower. Have a good idea? With almost no investment you can try it out. The signs of a market ripe for a lot of small innovations. Maybe we will have thousands of aggregators, all catering to small niches. Why not, indeed? And while these innovations happen, the classified site following the aforementioned strategy is stuck with essentially the old model, this time with audience scarcity created artificially in the ranking, while simultaneously giving up all other differentations towards other sources of classified ads.

Another option would be the move the value away from exploiting the scarcity of audience and/or attention and create it with something else. What is clear is that the prices for an ad are often marginal for the advertiser anyway, thus there is still a lot of money on the table, even a chance to recapture what was lost in the dwindling print market. What are the possibilities?

Service comes up as natural example. Maybe service in publishing the ad, making this ridiculously easy ("just send in your item, we'll photograph and describe it and send you the money when its sold". There are companies doing this for ebay). Make great ads, compete for highest conversion of lookers to buyers. Or good integration with backend systems. But most interestingly by handling the life and afterlife of the ad. Many types of ad are just there to look for the first and only customer that is willing to pay. All other leads just cost money if not properly redirected. The used car salesman will probably take any lead anyway, but if you're renting a flat for a predetermined rate, any affluent enough tenant is good enough and showing the flat to 20 people is only a cost factor. Find ways to qualify leads, automatically take the ad away when a good lead is found and maybe even monetize surplus leads in novel ways? That would be something that wasn't really possible offline!

A good neighborhood would be another possibility. Become an umbrella brand for trusted/valuable/.. ads. Become a label! I remember an old statistics of isbn.nu a price comparison engine for books. While amazon.com almost never had the cheapest offer (rather, they were in the middle of the field), they still got the most sales - by a large margin. Sure, many factors specific to e-commerce retailers play here, but one could imagine a brand for ads, established a firm standard for who can advertise, that translates into a trust that boosts the quality of leads significantly. Or have a trust system specific to your site. Ebay and craigslist lead here in their specific ways, but not much is seen here from traditional players.

In all these models, the increased circulation is really a blessing and the sites would be well served by making the work easy for the aggregators. Publish the ads in RSS feeds, maybe even additional APIs. By using something like feedtree you can offer an up-to-the-minute current feed at virtually no additional cost. Enable the market of new and hopefully innovative aggregators.

In fact, lowering their barrier of entry even further helps level the market lest no one single aggregator will come powerful enough to demand money for very current crawls or something like that.

It also enables a forest of mini sites, tailored to very specific audiences. A single parent community website maybe, that marks out neighborhoods with fellow members? An exclusive programming languages website, happy to point to job offerings for their favorite language.

To wrap up, just a quick example of how such a strategy could help a whole industry get out of the lock of a near monopoly comparison site: In Switzerland, every year the new rates for health insurance are published on the same day. Many, many people flock to comparis.ch, the leading comparison site. Their lock on the market is so strong, that they can demand high fees for every lead and dropping out of their engine is not really feasable for an insurance company. Although some smaller and actually cheap ones are actually not included, but few people know. One of the larger insurance companies, CSS, built their new price calculator around web-services. What if they would open it up, establish it as a standard and most other companies would follow suite? Creating a site more powerful and more complete than comparis would be of the order of an assignment for a CS student. The lock of comparis could quickly crumble, saving the insurance companies the fee for the lead, maybe giving it back to the customers.

While probably a bit harder to see this in the classifieds market, my guess is that a healthy transformation of this market could be driven by similar patterns; and surely there is still a lot of money there, you just have to figure out how to create value worth it. In any case, we are entering a phase of real innovation in this field, with interaction models and business models beyond almost direct translations of the offline world.

Jump Starting Recommendation Engines with Tagged Bookmarks

A couple of weeks ago, I came across an interesting post on tagging at topix. Especially the quote on Raw Sugar, about "value added search around the tagging done by individuals on their own data" got me thinking. There is certainly the way My Web 2.0 is integrated into Yahoo! Search (try it out, they import del.icio.us bookmarks and it's amazing how often I search for stuff I already bookmarked).

But even more intriguing could be document analysis. Services like Findory are based on finding documents similar to what the system learned were other interesting documents (grossly simplified), and the most important input for the known interesting documents is the users' previous click flow. While a unstructured bookmark collection would certainly serve as starting vector, a tagged collection could be much more useful: My tagging reveals what aspect of the article interested me. I might have bookmarked this site on paper airplanes, but the tag fun will reveal, that the reason was more a meta aspect of the document rather than a vivid interest in aeronautics or handicrafts. Tagging an article on apple with innovation will emphasize the portions on their methodology as what interested me (in contrast to stories maybe about users of apple products in general).

How would that be implemented? All these personalization engines will extract document features and look for repetitions among the pool of interesting documents (again, grossly simplified). Common document features will be emphasized, rare ones dropped. By grouping documents by their tags, the features that these documents have in common could be emphasized even more. Then we can see if other users tagged the same group of documents and again filter document features against their profiles for the tags they chose for these documents (with bonus points if the tags are actually the same, not just the tagged documents). From research in collaborative filtering, we know that it works astonishingly well as long as the prediction stays in the same domain: That we both read the Hitchhiker's Guide doesn't mean we like the same music. Tagged documents might be the element to effectively use collaborative filtering techniques to extract relevant document features.

Maybe someone like Findory will try this? While the target audience for this probably isn't larger than the web 2.0 crowd, it would surely be interesting and lessons might be applicable to other loosely structured collections.

Update 2005-12-13: Yahoo buys del.icio.us. Of course.

semi annual state dump

Yet again, it has been a too long time since my last entry. Let's see if I get a handle on this blogging thing this try. A lot of happened in the meantime. For one backup brain motivation for blogging almost completely goes into my del.icio.us bookmarks. Increasingly, I am become aware of being read there, too and start to try to be more interesting in my bookmarking - but not always.

It has been more than a year, since we launched map.search.ch. The technology, or rather the movement behind its increased usage, has since been christened AJAX and after Google last February, Microsoft this summer, Yahoo yesterday launched their version in a similar style. I'd say the dynamic interface we did has been thoroughly validated in the market, which is very gratifying to see! And it's nice to see, that principles like strong back button support are becoming an standard part of AJAX development (Oh, I wouldn't ever claim that all this happened because of us. I'm sure that the technology and idea was ripe, but it makes me happy that we were to first to pluck it).

And of course, development on map.search.ch went on. It is documented in the (german) company blog at about.search.ch. Highlights include points of interest including live updated ones like free parking spaces, real time water temperatures (which was more useful in summer..), webcams and my favorite, the real time public transport schedules.

And as a special treat, map.search.ch got nice national recognition by winning the Master of Swiss Web 2005 award, and even by a large margin!

So much for the incomplete wrap-up of the last half year. I am currently very involved in a new project, which will shortly see the daylight. But now on to some stalled but hopefully more interesting posts!

Gurtenfestival from the Air

We recently updated some of the photographic material on map.search.ch. The Gurten, a hill near where I live, must have been flown over during the Gurtenfestival, a popular three day open air festival: map.search.ch/gurten?x=371&y=-25&z=1024! The colorful dots on the eastern side are all tents!