Uber PHP Programmers: Lets talk Latent Semantic Analysis

Status
Not open for further replies.

krazyjosh5

theres GOLD in dem tubes!
May 28, 2007
1,010
19
0
Alright. LSA is some serious shit. Anyone have any ideas on implementation of LSA in PHP? Any scripts out there to do it for me?

The wikipedia page on it shows the basic formula for it. It seems like someone could have already implemented this formula and boiled it down to a simple script API.

This is specifically regarding implementations in content spinners.

Also, I know LSA is kind of aimed at entire article analysis but couldn't this be applied to a sentence or paragraph level? I think if that was possible, one could LSA two sentences, figure if they match, and replace their content. Duplicate that and you've got quality autogenned content.
 


Man, this is the kind of thing that keeps me awake at night! Not that I know enough about it to do much for you, but I dream about this kind of stuff. It causes me to think in the kind of abstract terms that no one I know understands.

In the version I envision, it gives people things they didn't even know they wanted, until it magically pops up in front of them. Almost like reading their mind... "Yeah! That's exactly what I wanted. How'd you know?"
 
Well the semantic indexing project is written as C++ libraries and they have a detailed API out on it...damn it...not another project.
 
Well, correct me if I'm wrong, but isnt LSA just running your text through a system of equations? And arent those equations posted here:
Latent semantic analysis - Wikipedia, the free encyclopedia
...and countless other educational research papers.

Therefore.. wouldnt it be as easy as coding a formula?

The more I look into it, the more I fear that I'm sound like a noob but who cares, I'd still like to see some intelligent discussion on it.
 
Well, correct me if I'm wrong, but isnt LSA just running your text through a system of equations? And arent those equations posted here:
Latent semantic analysis - Wikipedia, the free encyclopedia
...and countless other educational research papers.

Therefore.. wouldnt it be as easy as coding a formula?

The more I look into it, the more I fear that I'm sound like a noob but who cares, I'd still like to see some intelligent discussion on it.

What you're looking at is a system for taking a metric fuckton of documents and finding relationships between them. This is something that google can do because they literally have warehouses full of data to compare, but will this allow you to start with nothing, punch in two sentences, hit go, and have php tell you that they mean the same thing? No.
 
Status
Not open for further replies.