I don't think anyone besides geeks would use something like that. Google says very few people use the filtering functions built into their search. You have to take it to the Monkey Island level first, perhaps, where there is a way to build a query? The average person won't take the the time to learn or remember a complex syntax for a search engine.
A powerful search syntax is a good idea. It could be something as simple as a frontend to another search engine's API. As far as natural language search, though that's the goal of sites like ask.com which have been around for a long time now.
Getting Watson to process a single natural language query in under three seconds with a reasonable degree of confidence was a huge technological accomplishment with today's technology.
I think it will probably be several years before the technology exists to be able to do the same thing with sub-second response times, on a much larger inconsistently-structured data set, with a higher degree of confidence, for millions of users.
It needs to understand that "jarin" is the username on a website called news.ycombinator.com, which allows comments, and these comments are different to the stories posted on the site.
But of course "Obama's latest comment on CBS news" means the president's comments on the CBS TV channel's news service.
An equally obviously "Brook's latest comment on Fukushima" would refer to a blog post by Barry Brooks on the Fukushima reactor problems on his blog http://bravenewclimate.com/
Oh, and that "sort-by:date" bit needs to understand the difference between comment date, published date and any dates referenced in the comment or post.
Easy!
The truth is that currently Google doesn't do too badly on any of those queries, without any special parsing or "understanding".
From my understanding of language parsing (one college class plus a little work lately) its an incredibly difficult problem that is NOT easy to solve. It always worked for something like Dungeon or Zork because the language was incredibly limited. There is a huge difference in the number of verb/nouns you would consider in Zork versus in a search string.
This is one of those suggestions which I usually decide to file under the sub-heading "hey, all we need to do is write some software which understands English!".
Right now this matches a definition of "hard" which is very close to "can't be done" despite whatever Jacques may say :)
I was aiming more for, "Hey, all we need to do is write some software which understands how to parameterize a query based on the location of a basic subset of English!"
The software I was speaking of, is purely meant to digest a few known signifiers. The one spot I do understand is trying to interpret "Louis Gray," into something meaningful. An example:"
On source today about query elements" returns "from:'source' date:'today' query"
I tried to search google images earlier for 'map of south america in 1836'. I got a few maps and a whole lot of bird pictures. Even after busting out some google-fu, I was still getting birds. Fed up, I tried the same search in Bing -- no birds. Still no idea what all the birds were about.
Are you sure you didn't search for 'parrots of south america in 1836'? That returns birds for me. 'maps of south america in 1836' returns some maps, of south america, from various historical periods around the year 1836.
I did the same search shortly after making the above post and no birds showed up that time. I'm not sure what was going on, but I certainly saw half a page full of birds on that search term earlier today.
A powerful search syntax is a good idea. It could be something as simple as a frontend to another search engine's API. As far as natural language search, though that's the goal of sites like ask.com which have been around for a long time now.