Some Interests that may be Thesis
topics (Watters)
Virtual Documents and Search Engines
Virtual Documents are documents for which there exists no
persistent version. That is, the contents and/or the links
are created dynamically. On the web, these documents
usually contain the result of a database query, result of a
text search, or are composed from a template or user
profile.
The research questions that need answering include:
- how to index this class of document
- what search algorithms can be applied
- how would one bookmark this type of document
Web Usability
The information
retrieval community have established the metrics, recall and
precision, for measuring the effectiveness of retrieval from
bibliographic and text documents, where effectiveness is from
user's perspective rather than system efficiency. These measures
can be used to compare techniques and interfaces, etc. No
similar metrics apply to retrieval from the web. Although
precision may have some meaning in a web search the notion of
recall does not appear to be useful in this context.
Can we come up with a construct validation of usefulness of
information measures? The
work involves investigation of alternate metrics, proposing
suitable metrics or a single metric, and evaluating its
usefulness.
Background would include IR course.
Labels
Filters using site labels are useful in a range of applications
from
protecting children from adult content on the web to
selecting news items for individuals. A new approach to
filtering is the use of Label Bureaus or databases of
site rating labels.
The questions that we are trying to answer here
include:
- can PICs and PICsRules be used to run a
heterogenously sourced Label Bureaus
- can a class of agent be created that automatically
generates useful labels
- can a Napster-like architecture be designed to produce
a community collaborative label bureau
- how can one guarantee the currency and authority of a
given label
Search Engine Query Improvement
Users still havve a great deal of trouble formulating an
effective qery and effective search strategy on the web. In
this project we ask the following questions:
- can we improve search satisfaction if we identify the
goal of the search first
- can we use drop down choices to improve the selection
of query terms
- can we use the log data to detect patterns in
search queries both for individual users and for
communities of users
- can we use attributes to improve the quality
of search query (eg. person names, organization names,
geographical clues, gender, and time clues)
Community and Social Filters and Personal Portals
Portals are organized views of subsets of the web. Corporate
portals provide easy access to relevant corporate information.
Personal portals provide easy access to commonly accessed sites.
The questions of interest here include:
- what is the model for collaborative portals that include
process as well as data access
- how do we generate dynamic portals based on community history
and personal requirements
- what learning rules or models can be used here
- how do know if portals are useful
Wireless Usability Studies
As we move from
laptops to
smaller mobile wireless devices for our information needs, we
need to study the effect of smaller display areas and bandwidth
on the usability of interfaces on these devices.
The work will include the development of prototype client
interfaces followed by a usability study.
Background would include web programming skills and,
preferably, HCI course.
Search Engine Algorithms
Search Engines on the web are well established.
Nonetheless, improvements in effectiveness are needed
either for add-ons to established systems or to provide
specialized search functionality to specialized areas, such
as medicine. Some areas of research here include:
- how to index for searching data in databases that may
be included in web pages created dynamically (see virtual
documents above).
- how to harness the log data to see patterns for
presearching or re-indexing for efficiciency
- how to use communities of bookmarks for indexing or
labelling
- can we use attribute analysis to improve the quality
of search (eg. person names, organization names,
geographical clues, and time clues)
Ranking search results for users
Given that we don't have much control over the big
search
engine algorithms or results one area to look for improving
the results from a user perspective is the post search
ordering of the results for the user. Some avenues of
likely research include:
- can we provide multiple ranking schemes for users
so
that either the pattern of user behavior or user selection
can reorder the results on the fly
- what ranking algorithms apply to this scale of results
- how do we include learning in these algorithms so that
they adapt to new web documents and changing interests of
the users
Java Ring applications
Come and see me if you have a novel application to exploit
Java ring technology!
ETC!!