Rouge

- Introducing multi-Rouge: an efficient way to measure multiple scores with same sequences (in particular multi-references) (https://github.com/pltrdy/rouge/pull/29)
- New flags for setting metric/stats from the command line (https://github.com/pltrdy/rouge/pull/25)
- Example scores updated (https://github.com/pltrdy/rouge/pull/21)
- New flag to ignore empty sequences (https://github.com/pltrdy/rouge/pull/18)

- We fixed an error occuring in ROUGE-L calculation in the context of sequences with multiple sentences. (see PR<https://github.com/pltrdy/rouge/pull/6>)
- Better Python 2 support by using common libraries:
- `six` to handle strings (https://github.com/pltrdy/rouge/pull/12)
- `io` to read files (https://github.com/pltrdy/rouge/pull/13)