Why skirts?
Fashion seems to be a natural use case for such analysis. As for the language complexity skirts are less challenging than for example dresses as the word
dress can be used both as a noun or a verb. Similarly, analyzing shirts or other parts of the wardrobe would require additional separation techniques to differentiate between male and female clothing and therefore skirts seems to be a neat example. Technically, following other products e.g. beer requires more computational power due to the large number of tweets about it.
How?
The Fashion Trends application uses many technologies. Core functionality is written in Python with the use of
Twitter Streaming API and
Twython module to collect the tweets. Data is written into
MySQL database each time a relevant tweet appears. Once a day the data is pulled from MySQL and pushed into
MongoDB hosted on
mongolab.com. API to query the aggregated data in MongoDB is also written in Python using the
Bottle framework. All is hosted on
pythonanywhere.com. Frontend interface uses
NVD3 re-usable charts library written in JavaScript which builds on top of
D3.js to provide the interactive visualization.
Contact me if you are interested in the full dataset.
See more at domajno.github.io/me