|
I'm interested in extracting named entities with Snowball. Eugene Agichtein has a great presentation on how Snowball does it at http://www.mathcs.emory.edu/~eugene/talks/dl00.ppt Can someone explain how the vector weighting of the 5-tuples work in Snowball.
For example the middle vector could be a vector with these weights: {<'s 0.5>, <central 0.5=""> <headquarters 0.5="">, < in 0.5>} Is each word weighted (via Log or TFIDF) according to the current sentence? |