4 comments

[ 2.9 ms ] story [ 40.3 ms ] thread
Not sure how accurate this is, but I really like this. Seems to be kind of the way I would expect it to be. Can you guys share what data went into drawing these conclusions?
We monitor a continuous online twitter stream of 10 million tweets a day and do Named Entity Extraction with best of NLP algorithms and models. Our info-graphics are aggregate over months of trending on Frrole.
It speaks more or less what has been buzzing in these countries.. And the distribution is cool too, as per my assumptions!
Just 5 countries, the title and the map was promising more.

On the method: what is the ratio of tweet that are unclassified? I do not believe your data to be accurate on the whole picture. For example, tweet on good restaurant is very frequents in Asia, but restaurant's names will hardly match some keyword.