Searching nested/hierarchical data sets

18 Jul 2011 - 11:24am
3 years ago
7 replies
1088 reads
Marjorie Kaye
2008

Hi Folks,

I'm working on a web-based business tool for which we need to be able to search for destination folders in a tangled labrynthine mess of data with many duplicate folder and file names.  Because of legacy issues, the relevance of similarly named folders will depend on their parent(s). Seeing names of Parent folders on multiple levels should also help some users refine their search terms to locate the desired destination folder.

Any suggestions for elegant examples of this, other than OS finder/explorer interfaces?  

Thanks,

-Marjorie

 

Comments

19 Jul 2011 - 1:05am
RMattB
2010
"...many duplicate folder and file names.  Because of legacy issues, the relevance of similarly named folders will depend on their parent(s)."

Have you considered trying to rationalize or merge these duplicate taxonomies? No visualization will truly make dirty data structure look clean.

It'd be great if there were a tool like Prezi (which shows containment Venn diagram-style), but would also show the full path / breadcrumbs of where you are. A graphical way to do that would be to push the parent containers out to the edge of the screen, but leave just enough space for each of their names, acting as something like 'Venn breadcrumbs'.

Matt

19 Jul 2011 - 12:05pm
Danny Hope
2008

On 18 July 2011 17:27, Marjorie Kaye wrote: > I'm working on a web-based business tool for which we need to be able to > search for destination folders in a tangled labrynthine mess of data with > many duplicate folder and file names.  Because of legacy issues, the > relevance of similarly named folders will depend on their parent(s). Seeing > names of Parent folders on multiple levels should also help some users > refine their search terms to locate the desired destination folder. > > Any suggestions for elegant examples of this, other than OS finder/explorer > interfaces?

How many folders? What’s the maximum depth of nested folders?

20 Jul 2011 - 4:02pm
Marjorie Kaye
2008

There is no maximum depth!  The maximum I've seen is 8. 

20 Jul 2011 - 2:42pm
usabilitymedic
2008
I agree with Matt, Marjorie. Even the most elegant UI will not help poorly organized data. In fact, I believe it could actually confuse folks more than expected. If the UI is familiar to the users, they will have certain expectations of the accuracy of the results, yet, the poor data structure will likely omit valid results from any given query.
20 Jul 2011 - 4:41pm
Marjorie Kaye
2008

Thanks for responding, Matt, Danny Hope and usabilitymedic. 

Love the idea of "Venn Breadcrumbs".  I'm pretty sure breadcrumbs are a big part of the solution here.

Usabilitymedic - I was assuming the desired result would almost always be returned, but hard to pick out from too many results. 

Of course it would be great to clean up the database.  The multiinational corporate behemouth of which I speak is in the process of just that - a process that involves beaurocratic intricacies of epic proportions, several years, and lots and lots of $$$$.

For this 6-week engagement, I've got to come up with a way for users to find the right place to put their file

More Context:  The interface is mostly a metadata entry form, and the Destination Folder field will have autosuggest, validate, and allow cut-and paste, which will take care of most users.  A deeper database search is, happily, one of the less frequent tasks users have to do, but allowing it is preferable to having people have to wait for an email or phone response to send a file. 

Also - these are very big media files.  "Sending" can be a 30-hour operation, and of course much of this work is time-critical.

Fun!

 

 

20 Jul 2011 - 6:28pm
Diana Wynne
2008

Take a look at various search results formats. I just migrated a huge wiki to Confluence, and have similar issues with legacy naming that I can't fix.

Confluence has 3 main mechanisms for showing page hierarchy besides breadcrumbs: their spotlight style quick search field segregates results by type (so pages are grouped separately from attached graphics). If you get 10 results that all say "home" you can hover for a tooltip listing the space (essentially the parent folder for the selected content).

The main search results is similar to Google: page name is big, excerpt is shown, and then a URL which is essential the path. I find this is more helpful than the superlong breadcrumbs for pages that are nested deep in the hierarchy because only some levels are named in meaningful way. There are filters for creator/modifier name, asset type, etc.

 

Also take a look at different ways of retrieving metadata about the selection/search: for example the way the Apple OS or Microsoft search offers time-based filters like Today, This Week, Last 30 Days. I'm not sure what the equivalent facet would be in your case, but there may be a filter like username or department that reduces the number of results to a meaningful set.

 

It can be helpful to see breadcrumbs/hierarchy stacked vertically rather than just run together--may be easier to parse if the folder names are long. Or use bigger fonts for higher levels.

so 

a blah>

             2   blah>

                            b blah>

 

or even

a blah>

2 blah>

b blah>

versus the more compact a blah> 1 blah> b blah>

Atlassian has a lot of funky UI, so it's funny to point to Confluence as an example of anything. But it happens to be the tool I'm working with most now.

Hope this helps. Diana

20 Jul 2011 - 7:50pm
Christopher Rider
2009

Might be interesting to apply some pattern matching to the folder hierarchy. If you can identify which levels in a particular sub-hierarchy represent "important" divisions, you could treat them differently.

For example, say in a 6-level section of the hierarchy, you can detect that a bunch of the level 4-5-6 folders are very similar sub-hierarchies. That might indicate that level three represents a major division within that section of the tree. You could then apply some different UI to the level 3 folders.

This would of course require some serious engineering, but it might pay off.

Syndicate content Get the feed