Mysql 全文搜索索引 - Match() / Against() 对不同的结果给予相同的优先级

Question

I'm using Fulltext Indices to identify similar column content. I noticed that the match scoring is not quite as I'm expecting.

In my table I store names of videogames. When I'm searching for "Resident Evil 5", all Resident Evil games will get the same score.

select id, name, 
    MATCH(name) AGAINST('Resident Evil 5' IN BOOLEAN MODE) AS score 
FROM game 
ORDER BY score DESC 

Output:

7   Resident Evil Revelations 2     1.7317759990692139
36  Resident Evil Remastered    1.7317759990692139
39  Resident Evil 5     1.7317759990692139
2   The Evil Within     0.7758325934410095

In my case Resident Evil 5 should have the highest score, but it will just set the same score for all games containing the word "Resident Evil". Is there any way to improve the scoring? I don't want to exclude the other Resident Evil games from the list, but give a higher score to Resident Evil 5.

Answer 1

The number 5 is not participating in the match, probably because it's shorter than ft_min_word_len ; confirm that with

show variables like 'ft%';

If I spell out the full word "Five", this gives me something longer than ft_min_word_len , and the query works as I think you expected it to. See this SQL Fiddle for an example.

Answer 2

First of all, look into the fulltext settings of your mysql server:

> SHOW VARIABLES LIKE 'ft%';

The output might look something like this:

Variable_name             Value           
------------------------  ----------------
ft_boolean_syntax         + -><()~*:""&|  
ft_max_word_len           84              
ft_min_word_len           4               
ft_query_expansion_limit  20              
ft_stopword_file          (built-in) 

You look for ft_min_word_len . As in this example, the default value is 4 .

To change that, if you want one-character words (like your number) to be searchable, you can set this variable by putting the following lines in your option file (usually my.ini ):

[mysqld]
ft_min_word_len=1

Then restart the server and rebuild your FULLTEXT indexes:

REPAIR TABLE YourTable QUICK;

Keep in mind this will increase your fulltext index quite significant.

This answer is based on the assumption, you're using MyISAM as table engine. If you're using InnoDB, the keyword is innodb_ft_min_token_size .