edward_rolf

>Microsoft did it's best to kill it about a decade ago

A decade ago I started learning C#. I did not notice this event. What was it in particular that they did?

>outside the skill set of most .net developers

But javaland has many of those devs? I'm really wondering if anyone outside of the Lucene team and Google even has a clue about how to build a modern search engine.

If it is true that .net devs are unskilled then this project is the perfect starting point for all those who wish to transcend from being web and business application developers making tons of money for their employers, to ML devs, thus (hopefully) securing for a while their hireability, since businesses in fact have realized they need to become smarter than their compertitors and that they need ML skills (and big data capabilities).

It's the perfect starting point because in this project, I've taken all ugly, wierd and awkwardness out of IR, because there is no place for such things in IR, because IR is not ugly, wierd or awkward, it's very, very straightforward. The code needed to express an IR system needn't be complex, at all. Half the fun of doing this project is showing everybody how easy the domain of information retieval is.

Hi, this is what I would like the dotnet core community * to start using instead of looking for the Lucene team to port several years old java code to core. We already have a search engine on dotnet core. It's called Resin. I'm only a few minor optimizations away, plus one or two features I'd also like to include, from having a robust and full-featured Lucene replacement, leaner, faster, with better compression and with readable code (sometimes even object-oriented). Time to bury Lucene.Net people. Come help out.

*) Is there such a thing? I mean, you're awfully quiet and compared to the java community, which I know exists, you seldom help out in open-source projects. What's up with that? Haven't found the right project yet? Search, i.e. information retieval and NLP, is about as about as interesting as it gets, at least to me. General AI will be bourne out of IR and natural language processing. Not interested yet? There's a really cool Trie and some pretty sweet tree traversing algorithms that can perhaps be optimized. Have a look at my poor-mans [Levenshtein automaton over in-memory trie](https://github.com/kreeben/resin/blob/master/src/ResinCore/I...) and the same [over a disk-based trie](https://github.com/kreeben/resin/blob/master/src/ResinCore/I...). Sentiment analysis comes later on. It's great fun!

Bigtable sounds like a massive undertaking. What is it that will eventually make you upgrade your data architecture? Is it the amount of data that is the main issue or is it something else?

Sure I don't see why not. This is meant to be a Lucene replacement. What are you using currently on bubblehunt?