Abstract
News overload has emerged as a growing problem in our increasingly
connected digital information era. With complex long-running stories
unfolding over weeks and months, young adults in particular are left
overwhelmed and demotivated, which leads to their disengagement
from politics and current events news.
This dissertation presents a method for the automatic generation
of metro maps based on news content obtained from user-specied
RSS feeds. Metro maps are familiar to most adults, and they are
intuitive visual metaphors for representing concepts which branch
and diverge, such as news stories. The method described performs
entity disambiguation and various other NLP techniques to extract
a set of topics (metro lines) from a news corpus which provide a
cohesive summary of its content.
The diculty of drawing unoccluded octilinear metro maps is a barrier
to their current utility in InfoVis. Therefore, this dissertation
also introduces a heuristic force-directed approach for drawing metro
maps, which is rened using multicriteria optimisations taken from
neighbouring literature in information cartography.
The resultant system is demonstrated using the RSS feeds published
by several popular British newspapers, and empirically evaluated in
a user study. The results of the study support the hypothesis that
metro map users demonstrate greater topic recall than users of an
equivalent RSS reader. Lastly, areas for future research are discussed,
followed by recommendations for the commercial development of this
and similar systems.
connected digital information era. With complex long-running stories
unfolding over weeks and months, young adults in particular are left
overwhelmed and demotivated, which leads to their disengagement
from politics and current events news.
This dissertation presents a method for the automatic generation
of metro maps based on news content obtained from user-specied
RSS feeds. Metro maps are familiar to most adults, and they are
intuitive visual metaphors for representing concepts which branch
and diverge, such as news stories. The method described performs
entity disambiguation and various other NLP techniques to extract
a set of topics (metro lines) from a news corpus which provide a
cohesive summary of its content.
The diculty of drawing unoccluded octilinear metro maps is a barrier
to their current utility in InfoVis. Therefore, this dissertation
also introduces a heuristic force-directed approach for drawing metro
maps, which is rened using multicriteria optimisations taken from
neighbouring literature in information cartography.
The resultant system is demonstrated using the RSS feeds published
by several popular British newspapers, and empirically evaluated in
a user study. The results of the study support the hypothesis that
metro map users demonstrate greater topic recall than users of an
equivalent RSS reader. Lastly, areas for future research are discussed,
followed by recommendations for the commercial development of this
and similar systems.
Original language | English |
---|---|
Place of Publication | Bath, U. K. |
Publisher | Department of Computer Science, University of Bath |
Number of pages | 116 |
Publication status | Published - May 2017 |
Publication series
Name | Department of Computer Science Technical Report Series |
---|---|
ISSN (Electronic) | 1740-9497 |