Inaccuracies in the Google Knowledge Graph

The Knowledge Graph was an exciting project: a search engine that no longer just exposed you to the information of internet but could actually summarise certain topics for you and present them on the results page. Of course, this would depend on the fact that you were interested in the most popular topic associated with that keyword. But now it looks as if Knowledge Graph is presenting inaccurate or outdated information as much as 1 out of every 5 times.

The Birth of The Knowledge Graph

Google launched Knowledge Graph last month as a feature format update on their international site (It still isn’t available on some of their regional sites). Its development had been dogged by technical problems as this is an entirely novel way of aggregating and compiling data. Some called it destructive to SEO and web traffic as a whole. People looking for information now don’t even have to visit websites any more. Google says though, that the real intention is to concisely explain a concept, hopefully leading to further investigation.

Graph vs. Wikipedia

According to a recent study by SEO company Conductor, Google’s Knowledge Graph has a 12% disparity margin from Wikipedia. It seems much as case of there being too many cooks, which is why it’s as high as 20% for high activity search trends. Because these have wider sources of information and can be dominated by specific focus if stories relating to that topic have been in the news recently. On the positive side though, only 4% of low activity search queries were inaccurate. This disparity was understandably ironic, given that Wikipedia and Freebase are Google’s main source of information for this graph. However, as Wikipedia is a user-edited resource, and that the entries can be somewhat transient, many questioned whether this is the best policy as it is.

Outdated Entries

The main problem seems to revolve around how often Knowledge Graph is refreshing their data to get the updated information. On average they catch up between between 2 and 5 days after a Wikipedia entry changes. This may not seem like a long time in practical terms. But considering the speed of information propagation on the internet today, that can be measured in seconds, that much time signifies quite a problematic lag. If you’d like to go have a look at the Knowledge Graph in action, make sure to use rather than as it isn’t yet available locally. Although doing this will negate Google’s new localised search results.