Forum Replies Created

Viewing 6 replies - 1 through 6 (of 6 total)
  • Plugin Author Chris Gorvan

    (@chrisgorvan)

    Hi miskebols, have you updated to the latest version of the plugin? It sounds like the same problem Li-An had, which was fixed in 1.0.1.

    I’ve just successfully tested this by setting my language code to ‘nl_BE’, adding two posts titled “Bartiméus” and “‘Avonds” and then searching for both “Bartiméus” and “‘s avonds”.

    Here’s a screenshot http://i.imgur.com/FlNbiOr.png

    Plugin Author Chris Gorvan

    (@chrisgorvan)

    I’ve updated the plugin and it should now be able to identify predictive searches, which will no longer be logged. Thanks again for the feedback.

    Plugin Author Chris Gorvan

    (@chrisgorvan)

    You wouldn’t happen to be using a predictive search plugin, are you? The query logger can only be triggered once during the search process, but a predictive search plugin would likely make multiple requests, which I hadn’t considered and definitely something I’ll need to take care of.

    I’m talking about plugins like Ajaxy Live Search that allow a user to type a few characters and have the website make suggestions.

    Plugin Author Chris Gorvan

    (@chrisgorvan)

    I’m guessing this was when /posts was loaded, either directly or when set to the homepage, as I found and fixed a bug with non-searches which affected ordering. That’s in version 1.0.2 and should fix the issue for you.

    Plugin Author Chris Gorvan

    (@chrisgorvan)

    Thanks for the feedback. I’ve just pushed version 1.0.1, which includes unicode support for search normalization, so any alphanumeric unicode characters should now be displayed. It will still remove non-alphanumeric, such as # and $.

    I’ve tested it using fr_FR locale, searching for “téléchargement” and “téléchargement musique” and it certainly seems to be behaving correctly. Let me know if this hasn’t fixed the problem for you.

    I’ll look into setting the excerpt length as an option in the next minor release.

    Plugin Author Chris Gorvan

    (@chrisgorvan)

    Thanks! It’s a start, hopefully it will prove useful and can develop into something more intelligent. But initially, the choices are based on click-through rates, the kind that is used to summarise ad and email marketing success: Rate (%) = Clicks (#) / Views (#)

    These are calculated on a per-search basis, so if a post has been displayed 5 times for a specific search and clicked 5 times it has a 100% click-through rate and is judged to have been useful. Any click-through rate greater than 0 will cause those posts to float to the top five search results. The only thing that can beat the click-through rate is if one specific post has been ‘pinned’ by an administrator, giving them a way to artificially float new posts, for example. There’s a little more going on, but that’s the primary method.

    All of the plugin data is stored in separate tables, starting with ‘sro_’ and they’re used maintain search terms and links to post ids. I’m trying to keep the data storage requirements as low as possible, so it’s normalizing and aggregating data.

Viewing 6 replies - 1 through 6 (of 6 total)