A couple of years ago I wrote about how I got myself — and some other things — added to Google’s Knowledge Graph. Search has come a long way since then, but not enough to realise that I’m not particularly notable.

Recently I was showing a colleague how to use Google Trends. He typed in my name and, to our collective surprise, there I was as a suggested search topic.

Google Trends suggested result for "Glynn Davies"
Google Trends suggested result for “Glynn Davies”

This surprised me for two reasons. First, I stopped getting a Knowledge Graph result — or so I’d thought — shortly after I first wrote an article crowing about it. Second, I hadn’t consciously realised the connection between Knowledge Graph entities and the pre-defined search topics in Google Trends.

So I checked and, sure enough, if you search “Glynn Davies” you get a handy little search disambiguation panel, showing me and my two almost namesakes: the US Ambassador to Thailand, and a British MP perhaps best known for an incident involving sheep and the absence of trousers.

Disambiguation of search results for "Glynn Davies"
Disambiguation of search results for “Glynn Davies”

An interesting observation here is that, in Google Trends neither of these two people are offered when you search my name’s spelling (two “n”s). Google Search, on the other hand, doesn’t disambiguate by spelling alone. Presumably the behaviour in Trends indicates that the algorithm can tell the difference, but chooses not to in normal search results because people often can’t.

How to add things to Knowledge Graph: a recap on Freebase and Wikidata

As keen semantic search enthusiasts will remember, Knowledge Graph originally drew much of its entity data from a publicly accessible database called Freebase. It drew a lot from Wikipedia too and still does, but for anything and anyone unable to meet Wikipedia’s notoriously troublesome notability criteria (like me), Freebase was an invaluable back route.

Eventually, Freebase was phased out and most of its data was migrated to Wikidata. I stress most because a lot of entities, including me, were not migrated. So I’m now featured in neither of the two primary sources of entity data, and yet I still have an entity result. Curiouser and curiouser.

And a recap on structured data

Even when I first got a Knowledge Graph result, I felt sure that simply having an entry in Freebase wasn’t enough. I’d also used a liberal amount of structured data, including the Person schema, to link my website to my Freebase entry and social profiles.

The implementation’s a bit different now, but basically it’s something like this:

{
    "@context": "http://schema.org",
    "@type": "Person",
    "name": "Glynn Davies",
    "url": "https://www.glynndavies.co.uk",
    "homeLocation": {
        "@type": "Place",
        "address": {
            "@type": "PostalAddress",
            "addressCountry": "United Kingdom"
        }
    },
    "sameAs": [
        "https://twitter.com/glynndaviesesq",
        "https://www.linkedin.com/in/glynn-davies/"
    ]
}

Because Freebase is dead and I’m not in Wikidata, I now only refer to my social profiles. But the thing is none of this data, nor the sources linked to, provides my age or place of birth — two bits of info Google clearly has.

My place of birth was an important marker for my earlier experiments with Knowledge Graph. Since the exact place is relatively obscure (I usually just give the nearest city,  Manchester), I made a point of entering it onto my Freebase listing. That way I’d know for sure that Google was taking data from Freebase.

How do I still have a Knowledge Graph result?

So, Freebase is deprecated and my entry wasn’t migrated to its replacement, Wikidata. I still have the structured data, and specifically the references to my social profiles, but none of these sources give all the information Knowledge Graph shows.

The only answer, then, is that while Google say that Freebase has been replaced by Wikidata, they are nevertheless still using the old data.