Hello everyone,
On Twitter, given a system that infers interests of English-speaking users from tweets they posted or with which they interacted, is there any method to evaluate the system's output without asking users to do it manually?
For example, the system infers the following interests for the Donald Trump's account: The United States, Hillary Clinton, President of the United States, Election, Bill Clinton, Russia, Barack Obama, CNN, Republican Party, Democratic Party, Florida, etc.
Thanks.
Based on your example, simple entity extraction, combined with frequency analysis would get you somewhere already. It might be a good idea to check out ConText (http://context.lis.illinois.edu/).
Well, Its not my area but you need to look for research in complex systems, network science, artificial intelligence, machine learning...
In case you have accounts of public figures, you could use crowd sourcing platforms and ask users to verify or indicate the mistakes in the system output. However, if the accounts belong to random users, I would advise against that due to privacy concerns, even if the accounts are public.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude