Hacker News Reader

Comment by duprat | Hacker News Reader

Provided you fit the eligibility criteria, I've had some success resolving this by creating a Wikidata element with the correct info (and picture), waiting for it to propagate to the knowledge graph API as a new entry (5-10 working days usually), then asking the knowledge graph team to merge the two entries: https://support.google.com/knowledgepanel/contact/posts_on_g...

View on HN

But what if you don't want your picture online?

You can create a Wikidata entry and mark it as "different from" the other folks that it's being conflated with. Then the Googlebot has the information it needs to do the right thing.

That doesn't usually stop it from picking an image from the image search results.

Yes, at some point she would still need to "claim" the "knowledge panel" for herself by complaining to Google. But having accurate info in the community-curated Wikidata graph (that's used by pretty much every Big Tech firm out there btw, not just Google) ought to ease that process by avoiding that "Knowlege Graph" entry being fed from less accurate sources.

Perhaps I'm using it wrong (I've long wanted to, but haven't yet seriously played with Wikidata), but there don't seem to be any Andrea Vassells listed?

https://www.wikidata.org/w/index.php?search=Andrea+Vassell

So OP would need to create multiple entries, for the wrong people too? And since Google currently displays this panel without it coming from a Wikidata entry, how does it then know that this new Wikidata entry ties up to what it wants to show; that the image it is showing is related to one of the other new Wikidata entries?

Wikidata doesn't look great. Tried searching myself. I have a pretty big web presence in general and all Wikidata has is a scientific paper I didn't write with one author with my first name as their last name and another author sharing my last name.

Well, it's worth noting that it's not intending to be specifically for publications, it's not like Google Scholar or ORCID for example.

The reason it's interesting / useful for things like 'knowledge panel' is perhaps clearer if you look at a more complete entry, such as for 'apple': https://www.wikidata.org/wiki/Q89 and something on using the SPARQL API, such as https://www.wikidata.org/wiki/Wikidata:SPARQL_tutorial or https://www.markhneedham.com/blog/2020/01/29/newbie-guide-qu....

As I say, I haven't actually got around to using it at all myself, but that's why I hold it in higher regard than 'not great'. :)

Use a fake photo as provided above, algorithmically generated. Or use a photo of Sergey Brin or Larry page, they deserve the credit for this "feature" eh?

that would be vandalism of wikidata

Citation: "I chose not to put a photo of myself online".

Assuming that:

1) "Andrea Vassell" is a pseudonym (so that Google won't confirm claim for panel) or author just does not want send any extra information to Google

2) There are multiple "Andrea Vassell" writers, fitting Wikidata eligibility criteria

The solution would be:

1) Find origins of wrong images

2) Create a Wikidata item for each author and link each item to pages, which contain relevant images with https://www.wikidata.org/wiki/Property:P973

3) It is a good idea to create a page, describing original "Andrea Vassell" starting with a definition in a form of "Andrea Vassell is an NNNan writer, publisher of non-fiction books ...etc...". This page may contain the images of books (so that final card may look like https://www.google.com/search?kgmid=/g/122rd6_8&hl=en)

Step-by-step example:

1) I was notified that Google Knowledge Graph mixes Russian historian Andrey Simonov and American economist Andrey Simonov (wrong page snapshot: https://web.archive.org/web/20201202182248/https://www.googl...)

2) I created https://www.wikidata.org/wiki/Q41802044, https://www.wikidata.org/wiki/Q103187106, https://www.wikidata.org/wiki/Q103378461 and improved https://www.wikidata.org/wiki/Q4419737 (as all of them are passing notability requirements)

3) After few weeks Google removed photo of economist from historian (https://www.google.com/search?kgmid=/g/1hb_dk0rq&hl=en), because this photo was attached to economist card https://www.google.com/search?kgmid=/g/11fyy5b4yb&hl=en

Disclosure: I'm a Wikidata editor with 10M+ edits (mostly tool-assisted, of course). Have no relation with Google, other than hundreds of unsuccessful attempts to communicate with a brick wall of Google's contact form.

Why on earth are we providing free labor to fix googles mistake though?

I get why we might to help the various Ms Vassell's who are caught up in this though, but why are we pretending that Google is being reasonable here?

Another important notice. Ignore all suggestions from comments to provide fake data. Of course there are checks on Google's side. It is trivial to detect that image was generated with convolutional neural network. It is trivial to detect that you have used already known to GoogleBot image of Sergey Brin. Google marks sources with many contradictory facts as untrusted and pessimizes them when calculating weights for Knowledge Graph. This is not a speculation, this is literally written in published patents.

Reply on news.ycombinator.com