| Bug #2560 | Relevancy and fine-tuning MySQL Full-text Search | ||
|---|---|---|---|
| Submitted: | 29 Jan 2004 9:51 | Modified: | 22 Feb 2013 16:26 |
| Reporter: | Clay Webster | Email Updates: | |
| Status: | Verified | Impact on me: | |
| Category: | MySQL Server: FULLTEXT search | Severity: | S4 (Feature request) |
| Version: | 5 | OS: | Linux (Linux) |
| Assigned to: | CPU Architecture: | Any | |
| Tags: | ufr | ||
[31 Mar 2006 17:14]
Freek Dijkstra
Though I'm aware that this is no voting system, and I'm no MySQL developer, I respectfully disagree with this feature request. MySQL is not a language analysis or search tool, but a very fast database (which does indexing very well when it comes to full text searches). I belief the whole point of having a ignore list and min_word_len is not to prevent noise in the search results, but to minimize the index table for performance reasons. I think any application developer is much more capable of tuning the search results, since the type of search queries users type and the result they get is very dependant of the application (what are you looking for?), the length of the query and the contents of the database. In short, it is already bad enough that MySQL has a per-server ignore list (while a single hosting server may easily contain databases with content in different languages). Trying to create a per-server synonym list would add to unexpected results. (I already saw enough "bug reports" today for MediaWiki because people could not find short words, and were confussed by that). The strength of MySQL is in creating fast databases, not in creating ontologies. Let's keep it that way.
[22 Feb 2013 16:26]
Sveta Smirnova
Thank you for the feature request. Currently you can achieve this goal using full text plugin.
Description: Relevancy needs some help and you say it's on the top of your todo list. Synonyms would help. How to repeat: Search for "United States Pants" when you've got an index full of matches for jeans and USA, but no direct matches. Suggested fix: It would be great to add a synonym dictionary file in addition to the stopword file. The file could have one and two-way associations. e.g.- "jeans => pants" "Documentation" <=> "Manual" "US" <=> "USA" "US" <=> "United States" etc..