Come on, Google. You’ve gotta do something about this Knowledge Graph situation. I’m all for quick easy knowledge, and believe it or not, I’m a fan of the concept behind the Knowledge Graph, but the execution just isn’t ready for primetime despite being at the center of it for going on two years.
It’s one thing to briefly show users vandalized information from Wikipedia (and I think “briefly” is a generous word in this case), but when you’re showing erroneous information for businesses for weeks (at least), that’s not going to sit well.
Do you think Google is doing a good enough job with the Knowledge Graph for it to be such a big part of the Google search experience? Let us know what you think in the comments.
I’ve been writing articles about Google screwing up in the Knowledge Graph for quite some time now. Rajni Singh, Associate Manager at analytics firm Mu Sigma, noticed this, and reached out to me last month pointing out an issue her company is having with it. Thankfully, this particular instance probably isn’t costing the company much business but still, you want Google to have your business information right, especially when its plastered all over the search results page.
Mu Sigma was founded by Dhiraj C. Rajaram. Wikipedia, which often uses the same info the Knowledge Graph does, even has it correct. When you search for “Mu Sigma Founder” on Google, however, Google tells you it’s Shailender Singh.
This is what it looked like last month when it was brought to my attention:
Notice that the “knowledge panel” on the right has Shailender Singh as a playback singer, which is accurate. What’s not accurate is Singh being the founder of Mu Sigma. Also not accurate is the fact that anything involving Singh would be on this SERP at all. In fact, everything on the page that’s not Knowledge Graph-related looks pretty much fine. Google is pushing down the old relevant organic listings of old in favor of its own “knowledge”.
Slow clap.
Checking back on the SERP on Valentine’s Day, we find that Google has changed the SERP. Unfortunately it’s no more accurate:
Once again, the organic results look okay, but all the Knowledge Graph stuff is still for Singh, only now we get a different Singh picture, and some different stuff in the movies and “people also searched for” sections.
Why hasn’t Google fixed this? For one thing, we wrote an article about it, but Mu Sigma also notified them using their feedback tools, which are specifically designed to help Google improve its Knowledge Graph info.
I’m not going to go through every error we’ve encountered with the Knowledge Graph here. There have been too many of them at this point, but we recently pointed out how it completely fails (and is not as good as Bing) at info that changes with news. We looked at the Detroit Lions hiring Jim Caldwell. It took a while for the Knowledge Graph to catch up with that.
Then, when Microsoft appointed Satya Nadella as its new CEO, The Guardian noted that Google fell behind on that as well.
It’s not just about Google not getting the news fast enough (it does just fine with actual news results, by the way). It’s really about how long it takes for Google to update the Knowledge Graph in general.
Let’s not forget the famous “gay butt sex” incident, where Google left vandalized Wikipedia info for the St. Louis Cardinals in search results for hours during the World Series when it was a highly searched query.
Hey Google, there are some flaws in relying on Wikipedia for your knowledge graph information… (cc: @rustybrick) pic.twitter.com/X1zUl3hQkW
— Ben Cook (@Skitzzo) October 28, 2013
Andrew Isidoro wrote an interesting post last week about Google’s methods for updating the Knowledge Graph. Noting that it draws from various knowledge bases that push info to its “knowledge panels” as shown in patents Bill Slawski has pointed out in the past, he mentions that Google needs two sources of info to verify against before it will insert data into a panel.
He did some testing of his own (you can read about his methodology in the blog post), and said, “We have long understood that the Knowledge Graph needed multiple sources of information to populate a panel for an entity, and thanks to patents we had an indication that two separate sources of information would be enough to influence results. However, seeing this (albeit rough and ready) experiment in the wild gives a solid sign that this may well be the case.”
He goes on to suggest that Freebase can be used as a means of user-generated content to pass info into the Knowledge Graph, but also that for breaking news and time-sensitive info, “the Knowledge Graph simply isn’t ready.”
“The process of becoming (or editing) an entity isn’t well-known and as such will hamper the ability for Google to keep its panels updated to respond in (almost) real-time,” he concludes.
We talked a little about it on Twitter:
@CCrum237 If you look at how Google/Metaweb structured Freebase; it was built to take in information from DBpedia and republish it.
— Andrew Isidoro (@Andrew_Isidoro) February 13, 2014
@CCrum237 Often creating a 2nd source within a Freebase profile, which is why so many people report Wikipedia as the main driver of the KG.
— Andrew Isidoro (@Andrew_Isidoro) February 13, 2014
@CCrum237 Indeed. It's worth noting that this's all based on observation albeit fairly thorough. It'd be good to have an official line on it
— Andrew Isidoro (@Andrew_Isidoro) February 13, 2014
@CCrum237 They should have engineered it better then… 😉
— Andrew Isidoro (@Andrew_Isidoro) February 13, 2014
Do you agree? Should they have engineered it better, or is it good enough to be such a major feature of the search experience? Let us know what you think in the comments.
