Please read up on database design. The whole point of a relational database is to establish basic entities, and then to populate those entites with specific records. The design you are describing is fundamentally unmanagable, and probably unbuildable.
You would for example set up a table:
"person".
You'd establish an id for reach person -- a simple unique number will do....this is why mySQL provides an autoincrement function.
Each person record
would include maybe, first name, last name, user handle, etc.
Then a table profile
Each profile record would have a unique id (this unique id is called a 'key'). You'd include also a profile type column: people, books, topics and organizations, etc
Then a table comments
Again each comment would have a unique identifier. It would also include references back to the PERSON who made comment and the profile associated with the comment. These references are called 'foriegn keys' ... keys related to other tables.
commentid
profileid
personid
commenttext
Then a table called commentrating. Get the idea?
ratingid
commmentid (Foreign key)
personid (Foreign key)
profileid (Foreign key)
rating
etc.
Your data is now stored in 4 tables, not thousands.
You can easily write SQL to answer these questions:
Which profiles have comments?
Which don't?
Which users have posted comments?
How many users have commented on profile 22?
What is the average rating of a comment?
Which user always rates comments higher?
Which profile has generated the most comments?
Which profile has comments which have generated the highest ratings?
Which user's comments have the highest rating?
Etc.
Isn't this the sort of data you'd like to extract?
Based on the sort of data structure I describe, you could write DYNAMIC SQL queries -- in one simple unchanging line --that would produce the answers to the questons above merely by changing a single variable (the record key you are looking for).
I DEFY YOU to achieve the same result with the thousands of tables structure you are describing.
What you want is thousands of RECORDS, not thousands of tables.