Named Entity Trends Originating from Social Media
Nigel Dewdney · 2012
There have been many studies on finding what people are interested in at any time through analysing trends in language use in documents as they are published on the web. Few, however have sought to consider material containing subject matter that originates in social media. The work reported here attempts to distinguish such material by filtering out features that trend primarily in news media. Trends in daily occurrences of nouns and named entities are examined using the ICWSM 2009 corpus of blogs and news articles. A significant number of trends are found to originate in social media and that named entities are more prevalent in them than nouns. Taking features that trend in later news stories as a indication of a topic of wider interest, named entities are shown to be more likely indicators although the strongest trends are seen in nouns.