Show HN: AI-Powered Web Collections

Hi! I've just deployed a new feature to my search engine called "Web Collections". They are a slices of the WWW index, subsets. You decide their size. They are created from the advanced query parser GUI that appear after you query the WWW index. From the GUI, simply give your collection a name and click "Create collection" and the result set from the current query will be appended to your named collection. A link will appear leading you to a search GUI that will let you query this new collection as soon as it's been persisted.

All collections can spawn new collections because all are queryable.

You may create collections from data that doesn't originate from the WWW index but from somewhere else by utilizing the HTTP API to create/append to/read from and query your collections, freely, right now.

There is one last feature to implement before my intended MVP is done and that is to be able to reference collections in the query language so that AND, OR and NOT set operations can be orchestrated across collections.

It's got the AI label because I create graphs of word embeddings (bags-of-characters, bag-of-word, bags-of-topics) and virtual vector spaces of clusters of documents and because it's a NLP framework of sorts. It's what I use to analyse and create new language models. Each training session creates a new model that enriches the one it was based on.

Here's a demo and a question because I'm of course curious: would this a be slightly useful to you had it been at full WWW scale?

http://didyougogo.com