Abstract:
The internet has become one of the most widespread platforms for information
exchange and retrieval as the number of news websites is increasing rapidly. During
the last decade, most of the major newspapers have developed web sites providing
news and other information. In addition, web-only newspapers have also appeared.
News aggregator is a good substitute for news sites like BBC news. Because news
aggregators can index not just the content of the BBC news but all other news sites,
giving it a huge advantage in coverage. On the other hand, news aggregators may
complement online news sites. Because news consumers incur costs (time and effort)
in searching for news that are important to them and also they will compare the
expected benefit from visiting a news site to the expected search cost, where that cost
includes becoming aware of the existence of the site and finding how to navigate it.
There are few news aggregators like Google News, News Look Up, Fark which
provide news aggregation facilities, but they are proprietary and there are privacy
concerns about the user along with the biasness of these aggregators. In order to
benefit more from the available information, the objective of the research is to
develop a technical framework, gathering online news and approach to recognize
most important latest news and display the recognized news items that society is
interested in without any bias. Presenting crawled news items in a way that it displays
the trending topics in society will increase the awareness of the reader. In order to do
that news classification and ranking is a needed.
News items for the framework will be gathered through (RSS) feeds. Gathered news
feeds will be stored and will be preprocessed. Keywords will be extracted from an
algorithm that can be worked with any language that has basic Morphological tools
for language processing. Category classification of the news items will be done using
a method that is based on the keyword extraction algorithm. Topic detection and
classification of the news items will be done to the category classified news items
using an algorithm that requires no corpus for statistics or training data. The ranking
of the news article, topic and source will be done using an approach which is based
on the virtual graph model. In the ranking process, similarity between articles are
calculated manually and it will be automated using the cosine similarity.
