The genius of the web is that it is flat – anything can be a link and anything can be linked to. But this is also one of its greatest limitations. When you come across the word Taurus on the web, you probably know immediately if it the word refers to a car, a bull or an astrological sign. Your computer does not. You know because you are adept at using clues from the surrounding text or because of the context of your current activity. Because computers can not figure out this context by themselves we must do that work for them. We learn to add “car” as an additional search term if that is what we are looking for. We accept the fact that after reading a review of a Taurus car we will still need to type Taurus into the search box on AutoTrader because we have no other way to tell AutoTrader about the review we just read or what we were doing before we came to their site. The list of frustrations goes on. If we discover an event online and then want to book tickets, or just get it into our calendar, we start from scratch by telling whatever service or application we are using what we want to do.

We are, of course, huge fans of the web. It is an amazing resource, but the more we use the Web, the more we become aware of what it can not do. Tim Berners-Lee, the creator of the Web recognized the limitations of the web very early on. He also realized that if we could add a layer of semantic information to the web it would become possible for computers do much more for us.

Tim’s proposed solution to this problem is the Semantic Web – a way of adding semantic information to hypertext links to make those links more useful to both computers and people. The most well developed effort to bring semantics to the web is Microformats. There are now Microformats defined for concrete things like calendar entries and contact information and more abstract things like social relationships. This is important work that holds great promise, but it is also excruciatingly slow work. First, a format for defining the semantics of a class of things, cars, movies, lists, currency, etc must be defined and agreed upon, then the producers of web content must incorporate the format into the content they publish. Only then can applications and services take advantage of this semantic information. There are some great early examples of microformats in action, but they still today represent a tiny minority of the content on the web.

An alternative approach would be to use artificial intelligence to derive semantic information from the ordinary text that makes up the majority of the webs content today. This would save us the trouble of defining a format for every class of object on the web and would make it possible to extract semantic information from the language surrounding any object after the fact. This approach has the advantage that it requires no effort to define the formats and no work on the part of the content creators to incorporate those formats in the content they publish on the web. Unfortunately, it is a very hard technical problem, one that requires a lot of processor resources and even then works only sporadically in very specific situations. It is also hard to imagine that consumers would have the patience to wait for a complete semantic analysis of every page they visited before moving on to the next page.

It is easier to add semantics (and the services that depend on those semantics) to links in the context of a specific site. Amazon has taken advantage of their knowledge of the objects on their web pages to create a lot of useful features. When you are looking at a movie on Amazon, they can show you every movie by that director. They can show you movies that are in the same genre. They can even show you movies that have nothing in common with the movie you are looking at except that people who liked that movie also liked these other movies, a surprisingly useful trick called collaborative filtering. Amazon can do a bunch of other things that make your life easier on the web because they know you. They can push you through check-out quickly and automatically ship your goods via your preferred carrier. They can reference your buying history to remind you of birthdays or to personalize product searches. Amazon realized several years ago that their ability to deliver useful services to consumers was more important than their identity as a book seller and they broadened their offerings to enable them to capture more of the consumer’s attention, further strengthening their services, and ultimately enabling them to grow their share of the consumer’s wallet. As great as this experience is for loyal Amazon customers, some consumers are concerned that they creating a dependency that may not be in their best interest. They worry that Amazon can never encompass all of the offers that they would find interesting, and they are nagged by the suspicion that the everyday items they buy at Amazon could be found more cheaply elsewhere. At the same time, they wonder why they can’t have the richness and utility of their Amazon experience everywhere they go on the web.

This is the promise of the Semantic Web. If all of the objects on the web included rich semantic details about those objects, everything Amazon does today (and more) would be natively available anywhere on the World Wide Web. Alternatively, if Artificial Intelligence advanced to the point that rich semantic data could be derived in real time from every web page using the processing power of a personal computer, all of these services would be accessible to anyone, anywhere on the web.

A year ago, Alex Iskold decided that neither approach was likely to deliver on this promise quickly. So he founded AdaptiveBlue to see if he could offer a useful set of services to consumers in the meantime. He took a pragmatic approach combining a lightweight semantic analysis and a narrow focus on a limited set of popular objects like books, movies, wine, and electronics, to see if he could 1) identify many of the objects we come across on the web and 2) anticipate and present a set of possible actions one might want to take having found that object. If you are reading a review of a particular wine, would it be helpful, for instance, to have other similar wines or other wines by the same winemaker available a single click away?

When we met Alex last October, we were surprised by how often AdaptiveBlue’s analysis was able to tease out semantic information about objects on the web and how useful it was to have information and services related to that object immediately at hand. We believe that AdaptiveBlue’s services will simplify browsing, create meaningful recommendations, filter information, and enhance productivity.

AdaptiveBlue works across a large portion of the web today. There is no requirement for individual site owners to add their own metadata because AdaptiveBlue leverages their own top-down approach to building a semantic web. However, we are particularly excited about the prospect of the AdaptiveBlue developing tools that allow users to build the semantic web from the bottom-up to fill in the gaps and correct the top-down approach when necessary.

We are also pleased to be in business with Alex Iskold. We closed on a Series A investment in AdaptiveBlue late last week. Alex is an experienced entrepreneur and a contributor to Read/Write Web. We have been consistently impressed by Alex’s opinions regarding the prospect of a semantic web and the market for lightweight web services. We are excited to be a part of the dialogue around Alex, AdaptiveBlue, and the Semantic Web.

Recommended in The Nature of the Internet