Autocomplete

[Rene]

Dear Wrensoft Team!

We have been testing the new V7 Zoom Search.

The Autocomplete function would be a nice feature to have, the current implementation is less than optimal though from our point of view.

Manual entry of word suggestions is quite tedious.
Activating the option for parsing the logfile that lists search queries that users entered makes no real sense either. This logfile contains words that are spelled wrong, words that do not occur anywhere on the website and words that should not show up in autocomplete suggestions (like "porn" and so on.)

What we think would be a better approach is to use the zoom_dictionary.zdat file as the wordlist for autocomplete by default.
This file contains a huge amount of words that were indexed from the actual website. So they all occur on the website and there is a much wider range of words compared to the searchwords logfile.

Obviously one could take the dictionary file, clean it from things that are not words, and copy that into the manual entry box.
This would have to be done every day though, since we index every day and there might be new words on new pages. And the manual entry box has a character limit that is lower than our zoom_dictionary.zdats content length. So thats not really a solution for us.

It would be great to hear from you about this.

Thanks!
Rene

[David]

At the moment the auto-complete function is limited to 5000 words / phrases. (As a side note: there is a bug we found with this a couple of days ago. Trying to enter 5000 entries will cause a crash. This has been fixed for build 1003 later this week however).

It is common for the dictionary to have 200,000+ words in it. Usually there would be a lot of words you wouldn't want in auto-complete. So filtering out a useful sub-set would be hard to do in an automated fashion. And impossibly tedious to do manually. An even bigger problem with using the dictionary however is that it doesn't contain any phrases. So you would never get a suggestion longer than 1 word.

Activating the option for parsing the logfile that lists search queries that users entered makes no real sense either.

When parsing the log file, any search words that returned zero results are excluded from the auto-complete list. So you shouldn't pick up any misspellings, etc.. unless your web pages also contain the misspellings.

However there are lots of things we will be looking at for future releases. Including,

• Ranking the suggestions better. So more popular searches phrases appear first
• Optimisation of the auto-complete matching function to return suggestions faster, which in turn means we can have a larger list of suggestions.
• Using more data sources.
  • meta keywords
  • page titles
  • a separate text file
  • phrases already entered into the recommended links module

[Rene]

Thanks for the quick reply.

Looks like a complex topic

Anyway, i think the reason why misspellings did show up on our suggestions list is because we use "substring search" on all searches.

So a misspelled word might be part of another word and will therefore yield search results (its in german):
- "Preiss" (wrong spelling of "Preis")
- but is part of other words like "Preissenkung"

Looking forward to the upcoming Zoom Versions!

Thanks again.
Rene

[David]

By default a search for Preis will not match Preissenkung.

Which does bring up another interesting point. We should probably be excluding from the auto-complete list any wild card searches we come across in the log file.

[chinkle]

Hi -- Thanks for this new version. I'm implementing it on our test site, and was very pleased to see this autocomplete function; however, I'm not sure how it's meant to work. I use a custom form and a custom template, and the zoom_autocomplete.js file doesn't seem to be getting included in the page to provide any functionality. When I reverted to the default form and template, I still didn't see the .js load. What could I be missing? The .js and .zdat files were both generated as a result of the indexing process, so I think my settings were correct (I had it set to pull data from the log file, and that seemed to work very well as the .zdat file contains good entries). I'm using this on a .NET site. Any ideas?

[forestgreen]

Hello Everybody,

We have been using V6 for quite some time and now we have just upgraded to V7. I have tried to set up our search page properly in order to use the autocomplete feature.
While trying to do this I have come across the following two problems:

1. Autocomplete does not work for me in Firefox browser, yet it works well in IE 10. Is V7 autocomplete incompatible with FF? In FF I just get the input text field and nothing else.

2. The drop-down suggestions box appearing in IE search page overlaps the text input box by nearly 100%. We find this less than optimal, because I can't see any more what I am typing. Can we adjust the position of the box? If yes, how would I do that?

Anyway, thanks for providing this useful forum for discussing such topics.
Looking forward to your ideas to overcome my V7 problems.
Thanks, Christian
(from Germany, Europe)

[Ray]

Update--

Since V7.0.1021, the autocomplete feature can now automatically include words from Page Titles, and Meta Descriptions.

Also many bugs etc. have now been fixed.

You can find the latest build of Zoom here:
http://www.wrensoft.com/zoom/whatsnew.html#windows