[Sigia-l] ROI/Value of Search Engine Design - Resources?

[Karl Fast]

> Could you expand on these last two points?

I will try. [No! DO or DO NOT...there is no try. Bugger off Yoda!]

This is a bit long.


> > 3. LIS has failed to capitalize on this. One need look no further
> >    than OPACs (electronic card catalogues) to see how badly they
> >    have missed the mark.
> 
> How so? Too narrowly focussed on Subject/Author/Title?

It is well known that OPACs are difficult to use. Many papers about
this. A good one is by Christine Borgman called "Why are OPACs Still
Hard to Use?" It's in JASIS. 1996, I think. Can dig up citation if
necessary.

OPACs suffer from antiquated interaction models. Consider how few
OPACs conform to modern search interface conventions. Amazon doesn't
forces you to specify how to search (by title, author, etc.), you
just enter some terms and then refine. Libraries still force you to
start by saying how you want to search, and then what.

BUT, I believe the real issue is that users do not understand how
the OPAC is structured, and that the structure that is there gets in
their way (when I say struture I mean the structure of information)

I did a study last year (unpublished, *sigh*) comparing how
university students searched Google and an OPAC. It was clear that
people preferred Google. They knew the web was messy, unorganized,
and often untrustworthy. They were not sophisticated in their
queries. Despite this they felt confident in (a) Google's ability to
find things and (b) their ability to use Google effectively (note
the agency they would give to Google).

When it came to the OPAC they admired the structure, but were
mystified by it. They said things like "Gee, I love how it's
organized by subject...that's handy, I wish the web was like that"
but then they said "Damn, I don't have any idea how these subject
headings work." 

As an information system, the web is innefficient, yet searching
with Google is effective (or at least, people perceive that it's
effective). On the flip side, an OPAC is highly efficient, yet
people find searching to be inneffective.

I have some lovely quotes from the transcripts. One woman compared
searching the OPAC and Google something like this:

  The Internet is for everyone. It's for smart people and normal
  people and stupid people.

  Even stupid people can search the Internet.


To her the OPAC was something for smart people only. That's pretty
discouraging for a social instutition which is interested in the
public good instead of private profit.

  
> > In other words, LIS has figured out how to create a semantic layer
> > and that this layer is the key, but they've done a lousy job of
> > making that layer useful through an intuitive interaction layer.
> 
> Can you provide and example of what you mean by "semantic layer"? 

By "semantic layer" I mainly mean stuff like controlled vocabularies
which define relationships between things (equivalence,
heirarchical, and associative relationships).

I tend to think of a system having several layers (this is my
personal model and some of it's a bit fuzzy).

Content Layer
  At the bottom is the content layer (the things). In some systems,
  like OPACs, the content itself is not actually in the system (it's
  on a shelf). Same with a lot of e-commerce sites. But in a
  full-text journal database the content layer is there.

Metadata Layer
  Above the content layer you have a metadata layer. This is a
  descriptive layer that describes the objects within the system. I
  think in most cases there is a one-to-one mapping between the
  metadata layer and the content layer.

Semantic Layer
  Then you have a semantic layer. This defines relationships between
  items. Here is where controlled vocabularies and classification
  come into play (which you can almost view as the same thing).

  I call it a semantic layer because I tend to think of it defining
  groups of metadata items through the standard semantic
  relationships. But if you think about this layer primarily as a
  grouping layer then semantic is too narrow. For example, different
  versions of an item or all items by the same author.

Representational Layer
  Then you have the representational layer. This is how the things
  and their relationships are represented to the user (not how they're
  stored/represented internally).

Interaction Layer  
  Lastly you have an interaction layer that defines how you interact
  with the representations: how you select, manipulate, rearrange,
  and explore the representations.


Librarians kick ass on the metadata and semantic layers.

They suck on the representational and interaction layers.
  


I suppose the cognitive scientist would argue for another
representational layer beyond that: the mental structure in the
user's brain.

Does that kind of make sense?


--karl




--kal