Infer information from Tweets. Useful for human-centered computing tasks, such as sentiment analysis, location prediction, authorship profiling and more!
We could be connected to a Twitter stream, or we could have previously collected a corpus. We then use an active learning algorithm to select instances that we would like a human to label, such as through uncertainty measurement or through another measure like expected error reduction or information gain.
Another selection criterion might be: even if we are 100% positive that a document is positive or negative, we might still only be ~50% confident that the document is subjective. Therefore, that document might still be useful for the oracle to label.
We could be connected to a Twitter stream, or we could have previously collected a corpus. We then use an active learning algorithm to select instances that we would like a human to label, such as through uncertainty measurement or through another measure like expected error reduction or information gain.
Another selection criterion might be: even if we are 100% positive that a document is positive or negative, we might still only be ~50% confident that the document is subjective. Therefore, that document might still be useful for the oracle to label.