TweetSentiments.com provides a snapshot view of the world's mood via a heat map from an analysis of Tweets. The site uses Natural Language Processing and Machine Learning technologies to analyse Tweets from around the world and then presents the results on a Google Map.
As well as a heat map overlay the map also geo-tags individual Tweets. Looking at these messages you can see that the analysis often takes individual words out of context. For example, 'war' seems to always be given a negative score even if someone Tweets 'war' in a message about playing a computer game.
However these individual errors should be negligible if the system is built on a large enough dataset. Iraq and Afghanistan seem to score quite badly on the heat map, which seems to suggest a degree of accuracy; although you might expect Iran and Zimbabwe to not be quite as content as the map shows.
And, in case you were wondering, Mongolia seems to be the happiest place on Earth. At least for today.
_____________
1 комментарий:
Any links to API examples as to how this site is using that graphical map? I noticed the tile layers are like : http://mlt1.google.com/mapslt?lyrs=kml:cuxBQ5a_YXcUHfRaDu4oPGuLSRduo1ZEA|kv:1&x=3&y=2&z=2&w=256&h=256&source=maps_api&hl=en
Отправить комментарий