Information Technology Reference
In-Depth Information
the same news story or topic from a large stream of text messages. As news sto-
ries evolve over time, they continuously group and rank news articles in order to
constantly reveal relevant and hot topics and to enable their monitoring.
News texts do not only report facts about what has happened but also reflect opin-
ions of involved entities such as persons or organizations. They serve therefore as
a valuable source of opinions and help users making their decisions based on them.
Depending on the news type the opinions are directed toward a wide variety of topics
or other entities. For example, before political elections news articles echo the politi-
cians' attitudes toward current election issues and influence the vote behavior. The
perception of products or services is a precious piece of information for companies
and often a key factor in a company's decision-making process.
As with events and topics, news aggregation services facilitate the search for and
exploitation of opinionated text. Manually finding and evaluating opinion-relevant
parts may be infeasible for users. Therefore, the detection of subjective text parts
and the classification of text into different types of opinions are crucial tasks in news
processing systems.
In this chapter we focus on news aggregation services to organize and analyze
news articles. Section
1.2
describes approaches for grouping news articles depend-
ing on events and topics. We start by describing methods to detect and track short-
term events in news streams. Then, we discuss the clustering of events into more
abstract meta-topics. News articles often contain citations that underline reported
issues. Therefore, Sect.
1.3
concentrates on the extraction and evaluation of cita-
tions. Section
1.4
covers services to analyze news material with regard to expressed
opinions. We present a news aggregation system that incorporates all introduced
steps in Sect.
1.5
and conclude the chapter in Sect.
1.6
.
1.2 News Aggregation Model
News aggregation systems like Google,
3
Bing
4
and Yahoo!
5
organize and present
news articles from a large number of sources in order to offer users a comprehen-
sive supply of information. The enormous amount of news material published every
day requires a continuous and suitable preparation. In addition to the standard cat-
egorization of news article into the main columns such as “Politics”, “Economy”,
“Sports”, etc., and sorting the news items by date and/or language news aggregators
apply Topic Detection and Tracking (TDT) techniques to group news articles related
to the same events. Enhanced news aggregators offer additional services based on a
deep analysis of the news sources and material. For example, Google assesses the
sources and categorizes the content as
opinionated
,
detailed
or
preferred by the user
.
We introduce a system that not only focuses on a high-level classification of news
3
4
5
Search WWH ::
Custom Search