tagucci/pythonrouge ? reverse-engineered prompt

Build me a small Python library that wraps the ROUGE summarization evaluation tool so I can score my generated summaries against reference summaries.

I want to be able to pass in either raw summary text in Python or paths to summary files, and get back ROUGE scores like ROUGE 1, ROUGE 2, ROUGE SU4, and optionally ROUGE L. It should support common options like recall only or F measure, stemming, stopwords, length limits, and the usual ROUGE settings from the README. Please include a simple example script for English and another for Japanese, plus clear installation and usage instructions.

Make sure it works both when summaries are given directly and when they are stored in the standard system summary and reference summary folder format. If anything depends on external ROUGE files or data, set it up cleanly and mention what the user needs to provide. If you need to look up the current ROUGE setup details online, go ahead and do that.