is the answer to the question:
If someone handed you $0.5B and half a freakin' island,
how would you revolutionize graduate education in the information age?
By Nir Grinberg, Mor Naaman, Blake Shaw and Gilad Lotan
Why Real-World Diurnal Patterns are interesting?
What are the equivalent
Extracting indicative keywords of real-world activity
* method is available through our API
Aggregated Pointwise Mutual Information
For instance, given 10 coffee terms (e.g. “coffee”, “starbucks”, etc.) and their hashtag form, $APMI$ found: “#frappuccino”, “iced” and “#barista”
Context Pointwise Mutual Information
Context Pointwise Mutual Information (cont.)
Stoping criteria: $ContentPMI$ score > 500, that is, each term has to share an average of more than 5 terms with the seed's context.
For instance, for the same 10 coffee terms $ContextPMI$ found: “#cappuccino”, “macchiatos” and “#sbux” (shortened for Starbucks)
How well did our methods do?
Manually Labeled Top Terms
Quality vs. Quantity
(From the New Yorker, January 31, 1977, Illustration by Dana Fradon)
Hurricane Sandy in NYC
What did we learn here?
Work in Progress...