Question

I need to implement a custom scoring system in my database design. My program is a system of tags. My site is a dating site filled with tags about a particular person. A person will search for potential dates by using tags only. Each tag is rated from 1 to 10 depending on the person rates themselves on a certain tag. Here is how it works:

Search query:
Search for a girlfriend or boyfriend

Tag              Rating
---------------  ------
Good-with-kids   5
athletic-figure  9
intelligence     8
cooking          10

In the query above, someone searched for all these tags and their ratings against my database that has a collection of tags where people have rated themselves on these tags.

My problem:
When i run the query i can match profiles that contain all the tags. If a user runs a search query I would like the system to capture users that did not match all tags. If a query did not match all tags it should be ranked lower.

Do you know of a good database layout to do this?

Answer 1

Assuming

• A table containing a list of tags

  tags (id, name)

• A table containing a mapping of users to tags with a weight/ranking attribute

  user_tags(user_id, tag_id, weight)

Query

select user_id, sum(user_tags.weight) 
from user_tags join tags on tags.id == user_tags.tag_id
where tags.name in ('cooking', 'intelligence')
group by user_id
order by sum(user_tags.weight) desc