The Semantic Web is a great vision that began almost ten years ago, but at this point, the failures outnumber the successes by a very large margin. In order to move closer to realizing the great theoretical vision of its founder, Tim Berners-Lee, it is important to understand the common pitfalls of semantic undertakings. Let’s examine the top 10 failures of Semantic Web projects and companies one by one.
1) Semantic Web is Anti-Era
[login]The Semantic Web world is like a lumbering, giant group of smart people stumbling around without too much direction or vision, while all around them, smart and savvy entrepreneurs are creating amazing social, content, and mobile projects and companies at a comparatively break-neck speed and a fraction of the cost. In the last few years the Web has seen a revolution where small and cheap projects spring up, rapidly succeed or fail (and if fail, try again, succeed or fail, if fail, try again, and repeat this loop until succeed) and move forward, while Semantic Web projects are bulky, complex in nature, and require lots of time, money, and brain power before they can be even brought to market. They resemble projects of the late 90s where it took millions of dollars to realize whether a company would succeed or fail, while Web 2.0 companies can still participate in the “fail fast” startup approach where it takes 3-6 months and a few thousand dollars to realize whether the project will be successful or should be shut down.
2) Too Much Work Before Getting In-Front of Clients
The Web has shown that, while project organizers (large company projects or new startups) have to follow their vision, they also need to pay very close attention to the feedback of their users in order to get a sense of the true user experience of the product. The “rapid feedback loop” approach has garnered much attention and is prevalent in theories of Eric Reis and Steve Blank; and has been widely accepted throughout the business world.
Unfortunately for Semantic Web, its projects are complicated and not easy to create. This increases the amount of time that has to pass before the initial customer feedback is even possible, also slowing down ensuing feedback loop iterations, ultimately putting Semantic Web applications at a user-experience and agility disadvantage when compared to their Web 2.0 counterparts, because usability inadvertently takes a back seat to the number of other complex problems that have to be solved before clients even see the application.
3) No Talent To Hire or Partner With
Remember how difficult it was to get a good Java programmer to work for your company or project in the late 1990s? You almost had to give away your first born, a signing bonus, and be forever sentenced to deal with prima donnas who did not have to play nice, and jumped jobs for larger salaries within months. That was the result of not enough supply of talent to go around for a very large demand of it. Right now a very similar situation is occurring in the Semantic Web world. Hiring a good NLP engineer (computational linguist) will cost you an arm and a leg and so will hiring a good taxonomist, ontologist, or a programmer who can work with Semantic Web frameworks and build the application. Not only that, but they know they are difficult to replace, so you are at their mercy if they decide they want to misbehave.
4) Few People Make Up the Brain Power; If They Leave, It Might Kill the Project
Not only are Semantic Web professionals more expensive, harder to find, and have more leeway to misbehave, they can just kill a project simply by leaving because the work they do is often too complex for someone to just come in and pick right up where the original people left off. If the original people leave, the business owner gets stuck with a heap of unreadable code or a bulky ontology that now no one can understand, and lots of wasted time and money. Semantic Web technology is also harder to integrate and sell for the same reason — it adds extra risk to the buyer since if the founders or creators leave, they get stuck with the best scenario of having to hire more expensive resources who will have a 3-6 months ramp up time if they succeed, and much more if they don’t and other people will have to be hired. The world will have moved on by then.
5) Semantic Web Gives Only Incremental Benefits Over Web 2.0
Even the great Sir Tim Berners-Lee famously proclaimed that Web 3.0 is meant to live on top of Web 2.0 and not replace it. What is quickly becoming reality is that the incremental benefit of Web 3.0 over the traditional web is seldom enough to justify adding cost, risk, and time horizons to projects. It is becoming reasonably good practice to prove the business model success without even touching the Semantic Web, and if the business model works, to build out the Web3.0 components over time. Starting at Web3.0 without considering the lessons learned from simpler models is often a key to failure; just too lofty, too fast.
6) Ungrounded, Unbalanced Community Without Checks and Balances
Have you ever attended a Semantic Web conference, meetup, or an event? It is full of smart but nerdy engineers and taxonomists, with a few crazy futurist whacko’s sprinkled in-between. And if you look around the room you might notice a glaring absence or minority of business people and entrepreneurs. Worse yet, the smart people in the Semantic Web community constantly want to work on challenging and sophisticated problems because many of them have graduate degrees and it is what they have been doing. Unfortunately, that is rarely what the modern Web is all about. The modern web is about having a vision, executing rapidly, cheaply, and taking the simplest path to the goal; not turn it into elaborate, sophisticated projects.
Of course, there is room for elaborate and sophisticated projects, but the modern Web is about decreasing complexity, while a tendency of smart engineers is to increase scope and complexity because it is interesting.
7) Bad, Unclear, and Undocumented Ontologies
Even if you are one of the good semantic engineers, or have hired one, working with an out of the box existing ontology can be a frustrating, confusing and awesomely time-wasting experience. The ontology projects themselves often prove to be too bulky and large an undertaking for their founders and have a number of defects, flaws and inconsistencies. An ontology, like anything else, has a learning curve before it can be effectively used, and might appear to work for some contained use cases during development, but may show to be completely ineffective when used as the backbone of information in an eventual product that is presented to clients.
Another option is to develop the ontology in-house. In-house ontology development is a funny thing. It will take a long time to create, cost lots of cash, and in the end will seldom be much better than the out of the box ontology, and very likely still not be even close to be part of a production-level product, which is a scenario that greatly supports point #2 in this list.
8) Slow, Unscalable Taxonomy Work
If the previous point did not convince you, I should mention the process needed to create an ontology. Before building out an ontology, people start by building a taxonomy, which is an ontology with less complexity. Taxonomies are made manually. For example, if you are working on a simple concept like the idea of a person who helps to look after children, you have to ensure that all its synonyms (babysitter, home sitter, homemaker, nanny, au pair, etc) are also in the taxonomy. That means someone has to enter each and every synonym manually. This just does not scale. It scales even less if changes are required because the changes are always
manual and a huge pain, and again extra money and time thrown away.
9) No Clear Answers With NLP
If you have high hopes for NLP, the other big part of the Semantic Web, try to re-read nearly all the previous points of this article because nearly all the pitfalls regarding extra complexity, cost, and marginal improvement still apply. NLP engineers are hard to hire, expensive to employ and very difficult to replace. Plus NLP solutions, while really cool and helpful when they work, never seem to fully work and people have to settle for marginal correctness that is hopefully acceptable within their business model.
10) Business People Don’t Get It or Want to Hear About It
Ultimately, no one cares if the project is semantic, smart, stupid, or something else. Stake holders care whether a business is successful in and of itself, or if an enterprise project justified the investment. Terms like “Semantic Web” or “Web 3.0” are blanket terms for a number of separate technologies and sub-movements like RDFa, ontologies, open web, cloud computing, semantic reasoning, linking, and opening up the web, and more. Whichever one of these terms “Semantic Web” may mean for people, at the end of the day it matters little, because the goal is success via the most simple and fastest routs and the Semantic Web rarely offers that in practice.