MYSQL and PHP search script with word list and links

Question

I'm working on a search form/script for my website.

For a start every word used in content on my website is stored in a MYSQL table called words and it looks something like this :

id |  word | title_count | content_count | article_count | photo_count | video_count |
---+-------+-------------+---------------+---------------+-------------+-------------+
 1 | hello |           3 |             1 |             0 |           1 |           0 |
 2 |  what |           1 |             4 |             1 |           0 |           0 |

The word and id fields are UNIQUE, the fields title_count and content_count are used to store how many time the word was used in a title or in a main content. The other _count field just say how many time the word was used in an article/photo/video.

I'm not sure if all these count fields will be useful for a search function, but thought it could maybe come in handy.

Then I have multiple linking tables to all the _count fields of the words table that look like this :

id |  word_id | 
---+----------+
43 |        2 |
 7 |        1 |
 7 |        2 |

These tables are called : word_link_title , word_link_content , word_link_article , word_link_photo , word_link_video etc. The id field stores the id of a article/photo/video it links to and the word_id stores the the ID of the word linked.

Now that I have all of that set up I'm kind of stuck. I don't really have any clear idea how to sort out relevant content based on all these numbers and the research terms.

Something that I plan on doing is a search result page with multiple tabs, one there it would show all the results of the search and then in other tabs it would be separated in articles/photos/videos.

I have no idea if I'm actually on the right path to get something working. Hope someone can help me..

Answer 1

It all depends on what you want, if you want to suggest results, you might want to look into some famous word search algorithms : Jaro-Winkler is good for short words and Levenshtein is good for short words in long text. You can also use PHP's similar_text function for better matching after these previous algorithms.

I posted a bunch of search scripts here (Jaro-Winkler) and here when I built a search engine project, if you want to check it out.

As for the different counts, why not ? You can use the number of occurences to favor some keywords. But careful which ones you insert into your database ! You don't want words like 'the' or 'it' or any other common word corrupting all of this.

EDIT: of course, this means the search will be in PHP, and will have a huge inconvenient of needing to select a large amount of keywords (if not all) from the database. I ended up with a maximum estimated search time of 0.04 seconds, on a database with over 3000 words. So it seems ok =)