Facebook Graph Search: A First Step Towards the Semantic Web
The world at large may not be aware of Tim Berners-Lee. The British computer scientist was the first ever to successfully achieve successful communication between a hypertext (HTTP) client and a server via the internet—something that happens every time you load a webpage. He also created the term World Wide Web, and defined how it would work and be structured. Fundamentally, he established the internet we know today: a vast series of documents (web pages) connected into a web of information with hyperlinks. In turn, this also affected the way search engines would work: crawl the web using these hypertext connections, store pages in an index, and then return the best possible results to a user based on a search query. This format has worked fairly well as Google and Bing both return relevant results in response to a search term like “apple.” But a significant problem remains. Your results still have an equal possibility of showing you pages about fruit or pages about iPhones. That is because, on a basic level, the internet in its current form may be obsolete. Facebook seems to be betting that it is, and Tim Berners-Lee agrees.
The main focuses of the World Wide Web are the pages on which information is stored; users, of course, are focused on the information, not the page. This is why for a term like “apple,” you may be staring at both Steve Jobs and a Granny Smith. A search engine can’t inherently know what a page is about. Web content is designed to be readable primarily by humans and not computers, so engines have to essentially guess based on the appearance of a key phrase in the web page’s code and the amount of links from other pages on the web. It then uses an algorithm to determine how relevant the page is for a given term.
Berners-Lee has addressed this problem with something he calls the Semantic Web. Rather than an internet full of connections between documents, he asks, why not make connections between pure data? This would create a web of information rather than a web of documents , readable fully by both machines and users. They could translate human, natural language queries and return accurate results based on normal speech. Additionally, the connections and interactions between users and content also become points of data that can be indexed and searched. These interactions would allow data to be linked thematically—or semantically— eliminating our earlier smartphone vs. produce problem, as the two would become distinguishable by user interactions.
When Facebook announced its Graph Search in January, the collective groans of the Internet seemed to form human words for a moment. Those words sounded a lot like “Google killer.” Facebook’s graph search allows the ocean of data at the social network’s disposal to drive a search interface based on connections between users and their self-provided data. Graph search leverages your friends’ likes as a metric for displaying results. Its other crucial component is that a user can type in natural language queries such as “restaurants in New York that my friends who are divorced like” and the engine will understand the context of the query and display where your lonely friends stress-eat. It’s a system close to Berners-Lee’s vision. The important metric of quality is not the appearance of a keyword in a page, but rather the endorsement of quality information given by trusted people within your personal network—those endorsements (likes) being points of data themselves. Further, all data in the ecosystem is both completely machine and human readable. Every piece of data you provide can be interpreted directly by Facebook while still remaining readable to other users. Berners-Lee must be a happy dude.
The current limitations are significant, but solvable. One large problem is that the like metric’—and really all Facebook data—are provided at the user’s discretion. Therefore, the data available to be searched may not capture all the information in the network and may not be fully accurate. This data is also only confined to the Facebook ecosystem, excluding most of the information that lives elsewhere on the Internet in web page form. As far as being a viable competitor to Google, I don’t think Graph Search truly is. Facebook does have the traffic and user base to change overall behavior on the Internet—it has proven that in its short lifespan as a company— but the point of Graph Search is not to compete head to head with a company solely devoted to search. I think it may be an attempt to get people thinking about what is possible with the web as a whole, and maybe it will start to shift opinion. Either way it is a bold attempt at achieving the Semantic Web vision, and its real impact will soon be seen.