CSA402
Lecture 12
Problems that
Adaptive Hypertext Systems
attempt to solve
Problems
- Lost in HyperSpace
- Cognitive overload
- Complexity of the search space
- Search-browsing
- Static hypertext structure
- Inability to cater for different users with different needs and
requirements
- Representation requirements
Lost in HyperSpace Syndrome
- Characteristics
- User doesn't know where he/she is in relation to other (related) information
in hyperspace
- User doesn't know how to access previously visited node
- Causes
- Bad GUI design
- Poor organisation of information
- No links to landmark sites
- No/poor organisation of access history
- User unfamiliarity with content/ organisation
Cognitive Overload
- Affects both hypertext authors and users
- Causes
- Massive information spaces
- Authors as "experts" are expected to be able to provide links to all relevant
information...
- ... so that users don't have to think about the information they want to
consume...
- ... but this means that authors have to anticipate all the different ways in
which their information might be relevant to different users
- Authors cannot anticipate all the ways in which their information might be
used
- Authors cannot know all relevant information that exists/might exist in the
future
- Worst-case scenario: link everything to everything else...
- ... but users will then face cognitive overload
- Every time user accesses a node with more than one out-link, she has to decide
which link to follow
- Hyperspaces don't give guarantees about connectedness/ completeness
- A user's interaction with an information base improves with familiarity
- The more complex the organisation, the harder it is for the user to develop an
accurate conceptual model of the site
- Users also face cognitive overload in environments where the interface is
inconsistent
- 7+/-2 rule
Complexity of the search space
- Browsing is a form of search
- Definition of the serendipitous browser
"I'll know I've found what I want when I see it"
- Advantage and disadvantage of hierarchical 'classification' systems (e.g.,
http://www.yahoo.com)
- User has to know where to find what she wants
- Document must be classified correctly
- Lack of link semantics - no guarantee that in general hyperspaces a link is
going to lead to greater detail
- The more complex the search space, the harder for the user to construct an
accurate conceptual model
- The harder for the user to construct an accurate conceptual model, the
greater the likelihood of the user getting disoriented
- If the hyperspace is too simple, it probably won't cater for the needs of all
possible users
- If the hyperspace is too small, we can 'remember' where everything is
- AHS has to try to create and support simplified conceptual models which
precisely cater for the (individual) user's needs
Search-browsing
- Searching and browsing are complementary tools for navigating through a
hyperspace which does not provide a semantic representation of its contents
- Normally, search is used to identify a node that is "close" to the required
node, if not the required node itself
- Browsing takes place so the user can understand the information content and
make informed decisions about which link to follow
- Sometimes also necessary, because user is unfamiliar with terminology, so
needs to locate nodes which will enable user to specify more accurate query -
compare to RF in IR systems
- If user knows where information is, or if user knows how to get to
information (from some known landmark), then search not needed
- If user knows what is wanted, but doesn't know location of document, then
search is required
- Usually, search is performed at a different location from which user is
browsing
- Search-browsing would allow user to search while browsing, and the system may
enable the user to follow a recommended path to the relevant node
Static hypertext structure
- Hypertexts are usually static
- Authors create hyperspace, and users traverse them
- Hypertext generally cannot re-organise itself by learning from users
- Users who want to create a more easily navigable hyperspace need to create
it, possibly by replicating existing resources
- Links cannot be modified to lead to more useful information, unless the user
is the owner of the node
- Compare to similar problem in IR systems - IR systems generally don't learn
that query Q can yield different subsets of relevant documents D1
- Dn that have been selected by different users, but instead always
yield the same results list
- We know, from IR, that an infinite number of information needs can map to the
same formal query. Therefore, unlikely that a results list R will
satisfy all queries which map to Q
Inability to cater for different users with different needs and requirements
- The problems described above lead to an inability to cater for different
users with different needs and requirements
- Next lecture, we will see how AHSs solve these problems
Different requirements of Hypertext and IR systems
- Ideal AHS:
- As users browse through hyperspace, the AHS adapts the presentation of
information and the organisation of hyperspace to the needs and requirements of
individual users
- A user model is implicitly required, but is usually missing in hypertext
systems...
- ... what about IR systems?
- Most hypertext systems do not directly support search. Why?
- Most IR systems do not directly support browsing. Why?
- In IR, relevant information is stored in single documents, or in passages
(because that's what's indexed!)
- In hypertext, relevant information can span multiple nodes
- User interaction
- Hypertext: follow link - display selected document
- IR: submit query - display results list
- Organisation
- Hypertext: network, in which arcs (links) reflect relevance
- IR: (at best) document clusters, in which clusters represent similarity. (At
worst) no organisation - docs referred to in results list are relevant to
query
- Degree of relevance:
- IR: real number
- Hypertext: ??