One of the problems with making Google search better is that much of its indexed content is unstructured data. It’s easy to do deep search when all your data is nicely labeled. What happens when it’s just a big pile?
I have found that even unstructured data search can be tamed to a certain extent when you use something I call “persistent metadata” — metadata applicable to any physical object, like date and location.
Take you as an example. You were born on a certain date in a certain place. Let’s pretend you are very famous and there’s lot of news about you. If I wanted to research you, I might try Google News for 2012-2016, when you were in school.
By limiting my search to your school dates, I’m setting a *contextual boundary* around the data I’m searching. I’m using persistent metadata about you instead of keywords to focus my search. Even better, the persistent metadata I’m using is understood by most search platforms. I don’t have to tell Google “I want to know about when Famous Person went to college,” I just search 2012-2016.
Many of my Search Gizmos act as a bridge between the human expression of persistent metadata (“Grandpa Fred’s life”) and the computer’s understanding of how its information is organized (1895-1968.) Biography Buckets, Contemporary Biography Builder, Obit Magnet… that’s what they’re all doing. I’m always excited to find another way to explore Web search using persistent metadata, as I did last night.
endoflife.date is a service which tracks the end-of-life dates for over 200 products, including iOS, Firefox, Python, etc. It also has an API which allows you to get a product’s version history. When I learned about it, I wondered if I could make a Gizmo for easier historical software searching.
If you try to search about a particular computer program on Google, your results will weight toward the most recent version of the software, which makes sense. If you want to search for an older version, you have to narrow your search using something like version number, which can cause you to miss differently-expressed results.
On the other hand, if you use the persistent metadata of date, you can use more general search terms and still get relevant, useful results.
I had a fun conversation with Curly yesterday and made VerSearch, which uses the endoflife.date API to look up software by version and then creates searches for that product’s name by the version’s active lifespan. For example, I looked up Drupal 8.8, which had a lifespan of 12-04-2019 to 12-01-2020. That renders a Web search with a specific focus and a short result list. If they’re not narrow enough, just add keywords to your search. Since you’re working in a smaller data pool, even general keywords can reap good results.
The authoritative data offered by authoritative sources includes persistent metadata. I’m convinced there are endless ways we could apply this metadata to Web search to makes its results richer and more disinformation proof.