Summy generates excerpts for your posts by applying various algorithms for automatic summarization extraction.
To improve system accuracy various NLP algorithms are used. The most hard to find or build are the stemming algorithms because they are build with specific language rules. If you find or can provide a stemming algorithm for another language please contact me.
The term score algorithms TF-IDF and TF-RIDF require an internal database with linguistics statistics. At least for now it's not included.