Quote: stemmer behavior depends on stemmer weight; i.e., how aggressively the stemmer combines words

Quotation Skeleton

For all three word sources, at both tight … the following indices … Clearly, these are all related to the weight … [Porter, M.F., Program. 1980] is a rather light stemmer and Paice/Husk [Paice, C.D., SIGIR Forum, 1990] a heavy … [Lovins, J.B., Mechanical Translation and Computational Linguistics, 1968] somewhere between. [For the given source, Paice/Husk as about 1/3 better than truncating words at 5 characters, Lovins is somewhat better than truncating words at 6 characters, while Porter is about 1/4 better than truncating words at 7 characters. For truncating, 6 characters appears to give the best balance between under- and overstemming.]   Google-1   Google-2

Copyright clearance needed for quotation.

Additional Titles

Quote: stemmers are up to a third better than simply truncating words at five to seven characters

Quote: truncating words at six characters is better than five or seven characters

Related Topics

Topic: information retrieval with an index (32 items)