How to use my AI categorization API to build programmatic SEO
I’m working on an API for real-time classification of data and I’m looking for beta testers to give us feedback on the speed, accuracy and documentation.
It enables developers to provide a set of documents (which can be a word, sentence or paragraph) and a set of labels they’d like to apply to those documents. The API will then return each document with the most relevant label.
It uses ML under the hood to measure the similarity between each document and each label to return the closest ones. If you’d like to read up more about how it does it, check out.
I’m interesting in testing two things:
- Response times - our aim is to keep it under 100ms for <50 documents
- Accuracy - how good you think results are for your use-case
Example SEO page with IndieHackers’ posts
There are many possible use cases but one that we’re seeing take off is programmatically building pages for SEO - this is the practice of generating pages automatically from existing content optimised for specific queries you want to target. Here’s an example of NomadList’s generated content for SEO (hint: @levelsio isn’t writing each one of these individually):
Typically, these might be pages that are automatically generated from user-generated content that reference some query or keyword with relevant content.
To test this, I took the top 11 popular posts on IndieHackers at the moment and some queries I’d like to rank for in Google as my labels. It then showed me each category that each post should be in. I would then use this to create a page for each query with titles and links to those pages. Check out the labels it applied based on the titles, they’re almost all bang on.
If I wanted to take this a step further, I might want to then extract some sentences from the articles that most relate to the query and use that as an excerpt. Unfortunately, most of the top articles at the moment have little content so I picked the one also about programmatic SEO. I ranked the paragraphs based on the keyword and picked the top result to use an excerpt. It’s pretty good!
Try it out
You could use for this a lot of other use cases too, including categorising e-commerce data, diversifying search results, reranking search results based on semantics and more. I’d love to get some feedback, so if you want to try it out I’m opening up a beta here: https://similarity.ai
For those joining now, it includes 3 million characters for free per month which is about 600 pages in a novel. If you need more, let me know and I can increase it for you. We’re also not storing any of your data that you send in or get as a response - it’s ephemeral and real-time.
curl 'https://api.us-east-1.similarity.ai/classify/zeroshot' \
-H 'authorization: ApiKey {{ insert API key here }}' \
-H 'content-type: application/json' \
--data-raw '{"labels":["Food", "Technology"],"documents":["Crust pizza","Macbook Pro"]}' \
--compressed
