This is a kernel where the lexical items are compared using a similarity
function rather than just enforcing equality. For example you could use LSA or
word2vec to get word vectors and then the lexical similarity function is cosine
similarity between the vectors. Similar to subset tree kernel but should smooth
over lexical variation.
Citation:
Bloehdorn, Stephan, and Alessandro Moschitti. "Combined Syntactic and Semantic
Kernels for Text Classification." Advances in Information Retrieval: 29th
European Conference on IR Research, ECIR 2007, Rome, Italy, April 2-5, 2007,
Proceedings. Vol. 4425. Springer, 2007.
Original issue reported on code.google.com by tim.mil...@gmail.com on 21 Jan 2015 at 8:21
Original issue reported on code.google.com by
tim.mil...@gmail.com
on 21 Jan 2015 at 8:21