Evaluation is limited to English-language datasets with Wikipedia-based corpora, leaving other languages, domains, and corpus types unexplored.

来源：paper:PDF p.9, Limitations, fourth item