The original ROUGE supports this feature [-b n-bytes|-l n-words]
Usage: ROUGE-1.5.4.pl
[-a (evaluate all systems)]
[-c cf]
[-d (print per evaluation scores)]
[-e ROUGE_EVAL_HOME]
[-h (usage)]
[-b n-bytes|-l n-words]
[-m (use Porter stemmer)]
[-n max-ngram]
[-s (remove stopwords)]
[-r number-of-samples (for resampling)]
[-2 max-gap-length (if < 0 then no gap length limit)]
[-3 <H|HM|HMR|HM1|HMR1|HMR2>]
[-u (include unigram in skip-bigram) default no)]
[-U (same as -u but also compute regular skip-bigram)]
[-w weight (weighting factor for WLCS)]
[-v (verbose)]
[-x (do not calculate ROUGE-L)]
[-f A|B (scoring formula)]
[-p alpha (0 <= alpha <=1)]
[-t 0|1|2 (count by token instead of sentence)]
[-z <SEE|SPL|ISI|SIMPLE>]
[]
-b: Only use the first n bytes in the system/peer summary for the evaluation.
-l: Only use the first n words in the system/peer summary for the evaluation.
How to restrict summary length in pyrouge ?
The original ROUGE supports this feature [-b n-bytes|-l n-words]
Usage: ROUGE-1.5.4.pl [-a (evaluate all systems)] [-c cf] [-d (print per evaluation scores)] [-e ROUGE_EVAL_HOME] [-h (usage)] [-b n-bytes|-l n-words] [-m (use Porter stemmer)] [-n max-ngram] [-s (remove stopwords)] [-r number-of-samples (for resampling)] [-2 max-gap-length (if < 0 then no gap length limit)] [-3 <H|HM|HMR|HM1|HMR1|HMR2>] [-u (include unigram in skip-bigram) default no)] [-U (same as -u but also compute regular skip-bigram)] [-w weight (weighting factor for WLCS)] [-v (verbose)] [-x (do not calculate ROUGE-L)] [-f A|B (scoring formula)] [-p alpha (0 <= alpha <=1)] [-t 0|1|2 (count by token instead of sentence)] [-z <SEE|SPL|ISI|SIMPLE>]