By Andrea Volpini, WordLift
Search engines are evolving. They went from simple keyword matching to natural language processing. Those search engines got better at interpreting long tail queries. Yet they are still error-prone. Let’s put this into context: Google itself claimed that it started to use machine learning to cope with the 15 percent of searches done by users, on a daily basis, that have never been seen before. These are roughly 500 million new searches every day where Google doesn’t have a clue.
Semantic web is making those computational search engines taking a giant leap forward in three ways.
Answer the Unanswerable
First, the semantic web is helping machines to answer what before was unanswerable. Indeed, computational search engines uses context and relationships to give us answers that just few years ago were unthinkable (have you ever wondered how to write “2017” using Babylonian numerals? Now you can). This transformation is happening on the foundation of a W3C standard known as linked data. Wolfram|Alpha, Google and the other computation engine are using this vast amount of structured data to model their own Knowledge Graphs. These large interconnected web-scale databases understand entities such as people, events, topics, places and, most importantly, the connections between them.
The semantic web is working as a sort of check and balance in a lot of queries. Statistically driven algorithms like Google’s PageRank can easily fail or be artificially manipulated by linking a particular word or phrase to a specific website. On many long tail queries where the search volume is low it is hard to provide a solid answer just by looking at back-links and click-through rates (CTR). The result, in some cases, are the so-called Google Bombs. In these specific context the knowledge graph comes to the rescue.
Knowledge graphs have been created using the semantic web technology stack. The are built upon context, sourced from multiple datasets and verified with deductive logic. Thus, a knowledge graph, it is not only more expressive but it is also much harder to manipulate. A web that leverages on backlinks, user behaviours and semantics is a new kind of web. In this new web users can find better sourced information.
Everybody can build the web
Second, the semantic web also makes each individual contributing to its creation. Think about how Facebook is building a knowledge graph. The main purpose of it might be to offer a better navigation to its users and improved advertising options to brands. Yet Facebook is making users populate its knowledge graph with their sensitive information. And we don’t know how Facebook is really using it. Instead, semantic web tools like WordLift allow people to build data and keep control over it. Starting from a blog post. Yes this blog post.
In short, in the past decades the web got shaped through a top-down approach. Few minds created a bunch of algorithms that shaped the Internet. Those few minds influenced the lives of billions of individuals. Yet tools like WordLift are designed to democratise metadata and knowledge graphs. Finally the web gets back in the hands of everyone. Top-down algorithms and bottom-up knowledge graphs (created by people) can finally build the trustworthy web that was envisioned by Ted Nelson. (See more about Nelson on Wired article on Ted Nelson and Xanadu)
From search engines to semantic reasoners
Third, we nowadays give for granted the way the web evolved. Thus, it comes very hard for us to imagine a world where machines “reason” based on the input we provide them. Yet the semantic web is built upon the premise that machines can deduct answers based on the fact that we create. It’s called deductive reasoning (ie. having a statement like “John is married to Sue” as input we can infer that “Sue is married with John” without providing this specific statement). In simple terms we’re moving from search engines that crawl a web of pages to reasoning engines that infer knowledge from a web of structured data.
Although none of us knows how the future will unfold we do see a trend shaping the web. Semantic reasoners and computational engines like Google Knowledge Graph, Wolfram|Alpha or the Microsoft Sartori are about to create a new kind of knowledge using both deductive and inductive approaches. Most interestingly we foresee that there will be no a single major platform as we have seen with Google in the web of pages.
A glance into the future
Semantic web is inherently open and democratic. Anyone can contribute by linking his/her data with others’ data. Everyone, just like in the web of pages, can help building super-human knowledge graphs.
This revolution is starting from a single blog post. Articles like this one use AI to translate the text into structured data and to link this data with large knowledge graphs. Those knowledge graphs give birth to a new ecosystem based on machine reasoning at service of the human collective consciousness.
Andrea Volpini is the CEO of WordLift, co-founder of InSideOut10 and director of Insideout Today (an Egyptian award winning digital agency). He has 20 years of experience in online strategies and web publishing. In 2013 Andrea Volpini kick-started RedLink GmbH, a commercial spin-off focusing on semantic content enrichment, artificial intelligence and search.