Re: Solomon''s curse and search Bias

It's very interesting to see your work in some of the areas I've 
wrestled with.  I'm going to be studying that reference on query 
relaxation, which may be useful to me in my current work (which is not 
RDF-related at all).

More inline ...

On 3/4/2019 7:58 PM, Sherman Monroe wrote:
> Hi TomP,
> 
> Thank you for the reference to your work, I look forward to learning 
> about your ideas and perhaps they will help improve our interface.
> 
>     Another point to keep in mind is that a good user interface must be
>     very
>     different for a small project as contrasted with one that encompasses a
>     large collection of data.
> 
> 
> This is something we encountered earlier on. We wished to allow user to 
> group the result set by a certain field <http://bit.ly/2IQaC4o>. The 
> first prototype used a drop-down menu to store the fields list for user 
> to select the grouping criteria, but there are many fields (pages of 
> them). So we instead reuse the fields list. When a field is selected, an 
> icon appears adjacent to it and when clicked, the records list is 
> grouped by that field.

One of my experimental solutions, and I think it came later than the 
version described in my paper, was a ordered clickable list of all 
top-level subject terms.  You could expand the entire list to show all 
the following levels, but no one would usually do it because the entire 
list  was too long.  I added the ability to expand just a region of it 
by dragging the mouse over it while holding down the control key. 
Dragging while holding the shift key collapsed any section that was 
under the drag.

This was effective for me, but of course it is a pretty idiosyncratic 
technique that might not come easily to everyone.

Another method that worked well for me was a text search that returned, 
in separate tabs, hits on subject terms and hits on resources (i.e., URL 
page titles).  Clicking on a link in the subject term opened a panel 
showing that subject term and related ones, including some context. All 
items were clickable, of course. This provided a fast and easy way to 
explore related subject terms.  The display of related subject language 
terms was the probably the most successful part of my experimental 
program, as I have used it over the years.

>     The kinds of things you are trying to explore can also be approached
>     using a framework of library science, and specifically the concepts of
>     "navigation" and "collocation"
> 
> 
> Collocation is a genetic feature of RDF, e.g. when classes and 
> properties are viewed as facets of the subject that can be toggled. 

Yes, and it will be most effective when the RDF has a well-understood 
structure around each subject of interest.  Otherwise, how could the 
system know how far to go while including related information?  In other 
places, I have called these kinds of structures "idioms". One of the 
things I liked about topic maps vs RDF was that topic maps provide a 
built-in idiom template.  OTOH, that makes it more complicated to 
populate than RDF, where you can just throw in bits and pieces as they 
become available.

Of course now, with SPARQL, there is no comparison in querying RDF vs 
topic maps.  Back in 2002-2003, there was no SPARQL, and topic maps vs 
RDF was more up for grabs.

TomP

> Query relaxation techniques 
> <https://www.semanticscholar.org/paper/Answers-Partitioning-and-Lazy-Joins-for-Efficient-Ferr%C3%A9/488dff441f61c7c290ae0d2904a6fbc3c0f88392> 
> can also aid in serendipitous discovery. We feature a trivial version of 
> this with the Expand Search button and Remove Keywords button, for 
> instance, try removing the keywords from this list of movies 
> <http://bit.ly/2J5Rtvt>. The result is this nearby list 
> <http://bit.ly/2IR9mxC>. We plan to introduce features similar to this 
> later on.

Received on Tuesday, 5 March 2019 04:29:14 UTC