Module text_gen 

Text generation evaluation metrics

Provides BLEU, ROUGE (1, 2, L), and Perplexity for evaluating text generation, translation, and summarization models.

Functions

bleu_score

Compute BLEU score with modified n-gram precision and brevity penalty.

perplexity

Compute perplexity from log-probabilities.

rouge_l

Compute ROUGE-L F1 score using longest common subsequence.

rouge_n

Compute ROUGE-N F1 score (n-gram overlap between reference and hypothesis).