The Snowball project has defined stemming algorithms for 17 languages. Libstemmer provides these algorithms as a C library.

CL-LIBSTEMMER includes the full source of libstemmer, and will attempt to build and load libstemmer.so when it is first loaded. Obviously this will only work on a system with make.

The preferred way to use CL-LIBSTEMMER is with stem-all:

(libstemmer:stem-all '("visible" "irradiate" "vainglorious" "habitat")
                              :en)
=> '("visibl" "irradi" "vainglori" "habitat"), T

stem-all takes a list of words, a language (as a two or three letter abbreviation) and, optionally, an encoding. If such a stemmer exists, stem-all returns the stemmed words; otherwise, it returns the list of words unchanged. The second value is T if stemming was actually done.

You can also stem incrementally, using with-stemmer and stem:

(libstemmer:with-stemmer (stemmer :en)
  (libstemmer:stem stemmer "resplendent"))
=> "resplend"

There are also unbalanced load-stemmer and close-stemmer functions. Bear in mind that loading a stemmer is relatively expensive: for best results, stem in large batches.

Besides libstemmer itself, CL-LIBSTEMMER also includes the lists of stop words compiled by the Snowball project.

 (libstemmer:stop-word-p "is" :es) => T