A rough SQL implementation of tf-idf <http://en.wikipedia.org/wiki/Tf%E2%80%93idf> Assuming you have run the snippet <https://gist.github.com/sumanthprabhu/8066438> to generate a count matrix and stored it in the file 'tfidf.csv', the output will be a table named "results" containing normalized scores for each term per tag.

tfidf.sql

load local data infile 'tfidf.csv' into table tfidf fields terminated by "|" lines terminated by '\n'(term, tag, count);

DELIMITER //
CREATE PROCEDURE tfidf_applier()
begin
	declare res1 INT;
	set res1 = (select count(distinct tag) from tfidf);

	drop table if exists log_term_table;
	create table log_term_table(term varchar(200), logval decimal(20,5));
	insert into log_term_table (
		select term, log(res1/count(tag)) from tfidf group by term
		);
	drop index term_log on log_term_table;
	create index term_log on log_term_table(term);

	drop index term_tfidf on tfidf;
	create index term_tfidf on tfidf(term);

	drop table if exists tmp_results;
	create table tmp_results like tfidf; 
	insert into tmp_results( 
		select t1.term, t1.tag, t1.count*t2.logval as count from tfidf as t1 
		FORCE INDEX(term_tfidf) INNER JOIN log_term_table as t2 
		FORCE INDEX(term_log) on t1.term = t2.term
		);

	drop index term_tmp on tmp_results;
	create index term_tmp on tmp_results(term);


	drop table if exists results;
	create table results (
		term varchar(200), tag varchar(200), count decimal(20, 5)
		);
	insert into results(
		select t3.term, t3.tag, t3.count/t4.maxcount from tmp_results as t3
		 FORCE INDEX(term_tmp) INNER JOIN (
		 	select term, max(count) as maxcount from tmp_results group by term
		 	) 
		 as t4 on t3.term=t4.term
		 );
	

end
//
DELIMITER ;