Personal Information Retrieval Visualization (PIRV):
Clustering and Visualization of Web Document Search Results
Abstract
Conventional web search engines often return long lists
of ranked documents as their output. Since
only a part of the list of documents can be shown at a time, users cannot get a
complete picture of the returned documents in the first place.
Due to the imprecise nature of current Web search engines and the
explosive increase in the number of documents available, users are forced to
spend a significant amount of time going through the list of the results or
abandon the current search result.
In this project, we design and implement a system
called PIRV (Personal Information Retrieval Visualization), which dynamically
groups the search results into clusters and presents these clusters in
2-dimensional graphics. After
receiving a query from a user, PIRV sends it to the search engine, receives the
returned documents, clusters these documents according to similarity values
between individual documents, transforms the data into a graphical
representation, and then displays these graphics to the user. With this visual
display, a user may use his visual perception to evaluation these clusters and
to make an intuitive judgment about the relevance of these documents without
having to read a significant portion of each document.