APIHound Blog

Apr 9, 2015
Whoa! Long time since the last update. You didn't think we had gone stagnant like those other API sites, did you? Actually, we've been hard at work on the relevancy issue I wrote about in my last post. We've had a very computationally intensive analysis and mapping process running for a [trade secret] amount of time. This should have a very positive impact on search results quality once we integrate the output with our current API Index. I'll post an update here when it's ready.

Jan 5, 2015

A little bit about relavancy. Although we hammer on the weaknesses of hand-built directories (Yahoo! in the late 90's or ProgrammableWeb today), this super-manual approach is good at one thing: relevancy.

As you can imagine, that relevancy comes at a cost. It's impossible to manually scan the entire web, no matter how many employees you have. So the price for "better" content is that you leave a ton of material undiscovered. And then there's the ranking issue. How do you sort what you collect? Human-powered lists typically default to alphabetical order, which is weak.

In our attempt to build a better product, we immediately realized the challenges would be threefold: 1) Find all the APIs we could 2) Associate them with specific keywords and categories and 3) Rank the APIs within these keywords and categories. All this has to be automatic.

Many people might think heavyweights like Google have already solved everything that has to do with search. But indexing web-pages intended for human readers is not exactly the same thing as building an API search engine. For starters, the usefulness of concepts like PageRank do not apply well to APIs when calculating relevancy. Should REST APIs rank higher than SOAP ones? JSON over XML? What makes one API more relevant to a search term like "weather"? There are a number of approaches we can take, from using NLP (Natural Language Parsing) on a site to looking at underlying HTML metadata. This is a never-ending journey, and we have only taken the first few steps down the path.

You will likely see our results' relevancy shift as we experiment with different approaches. You will probably see non-API related material sneak in through the cracks now and then (I'm talking about you, American Petroleum Institute!)

Pointing out the weaknesses of our peers means we're going to take our share of licks as well. But we believe in the long run, an automated approach will yield the best product.

Dec 15, 2014

We're here.. but we're not exactly advertising that fact.There's a lot that goes into the launch of any new site: getting servers, configuration, app development, and (of course) putting together the UI that makes everything possible.

Our focus is on pulling all the pieces together and testing, testing, testing! We need to make sure all the pieces are in place so that we can begin to deliver the value we believe we have to offer.

If you spot an issue or have a suggestion, we'd love to know. You can contact us on our About page.