Database Query/Logic Question ?

[romeo0830]

Im looking to find out the logic which would be used to categorize millions of items with different keywords.

Ex.

Clothing | Jeans |

Matching similar items will match both keywords

Matching exact keywords will match keywords provided in the data base.

What would be another way of matching keywords without having either many mis categorized items or having to add thousands of keywords to the database?

Please help. my a i m is Michaelangelo374 or reply on here.

Thanks !

[Orjan]

it depends on how you want to categorize. If you only want one branch of categories, or multiple branches.

you could do categories as an own entity and make it recursive for parent categories
then you know all parent categories if you set one, like if you set "jeans" you can easily find it's parent "clothes"

if you would have only one branch of categories, let the child category be an property of the object, if you can have several branches (as you might also have a fabrics category for the jeans, so it's both "clothes/jeans" and "fabrics/denim" (and maybe even "fly/buttons" or "fly/zip") as categories, you need to create a joining table having object id and category id.

[romeo0830]

Thanks for the reply,

We actually currently use Parent Categories like you explained. Using joining tables would be good to narrow down items further but it still does not help the initial problem which is categorizing the items into the appropriate parent categories without having to add keywords to those parent categories.

Any other suggestions?

[WingedPanther]

I'd like to help, but without knowing what the categories are, how they relate to each other, and how you want to search, it's kind of hard to say. You may also want to create some functions in your database to facilitate the process... which will depend on the database you are using.

[Orjan]

Well, if you add the child category to the object, you don't need to add all the parents to the object, if you have created the categories recursive.

example of recursive categories:

id, name, parent
1, "Clothes", 0
2, "Fabrics", 0
3, "Jeans", 1
4, "Fly", 0
5, "Buttons", 4
6, "Zipper", 4
7, "Denim", 2
8, "Silk", 2

[romeo0830]

Hey thanks for your reply:

Would the above statement be more for narrowing down items?

Right now we are mostly trying to just get the initial categorization done.

Even trying to match Jeans with clothing we still have to associate some logic that will identify that jeans is a clothing category without having to add jeans to the keyword section.

For example:

Clothing: Jeans, Pants, Skirts, Dresses

Now all of these categories will go into Clothing but for millions of items it is not as easy to add all the appropriate keywords to there different parent categories.

Is there maybe some king of association program to match sub categories with parent categories.

See the way we are doing it now is ok, but when you have 600,000 plus uncategorized items it is pain to try to keep adding keywords to the main category so next time you update that item gets categorized along with maybe a couple of more thousand items. This is just a example that although we can keep adding categories manually it will eventually come to a point where it will be to much to do.

Any more suggestions?

[Orjan]

Well, the primary work to categorize the articles is never avoidable unless you already has something like a grouped article number or similar to choose from.

having 600,000+ items is many hours of work and actually, I believe there is no machine way to perform it. maybe do some searches on the article name and see if it matches on a few.

[romeo0830]

Thanks again for your continued support

Yea, thats what i was worried about. I guess a little hard work never killed no one but you would be surprised what you can live through. =)

Hey thanks anyway for your help, I appreciate it.

If anyone else has any ideas please shoot them my way. Thanks !

[Orjan]

That's the sad thing about computer systems that has been running for a while when you understand you want to get other information from the system which needs new data on the objects stored.

And for doing categories like this, rather sharp guidelines is needed if the job is splitted on several persons, because sometimes, it's not obvious what belongs to what...

[romeo0830]

Yea it is, well hey if you got A.I.M message me at MichaelAngelo374 we can talk some more.

Thanks Mike !