Score ROUGE
ROUGE, qui signifie Recall-Oriented Understudy for Gisting Évaluation, is a set of metrics used to evaluate the quality of summaries produced by automatic summarization systems. It is particularly popular in traitement du langage naturel (NLP) and is often employed to assess the performance of models that generate text, such as summarizers, traduction automatique systems, and other génération de texte outils.
ROUGE compare principalement le résumé généré à un ou plusieurs résumés de référence (souvent créés par des humains) pour voir à quel point ils correspondent. Les principales métriques incluses dans ROUGE sont :
- ROUGE-N : Measures n-grams (contiguous sequences of n items from a given sample of text). For instance, ROUGE-1 evaluates single words, whereas ROUGE-2 looks at pairs of consecutive words.
- ROUGE-L : Focuses on the longest common subsequence between the generated summary and the reference summaries, taking into account the order of the words.
- ROUGE-W : A weighted version of ROUGE-L that accounts for consecutive matches and penalizes gaps.
The scores generated by ROUGE are typically expressed as recall, precision, and F1-score. Recall measures the percentage of n-grams from the reference summaries that are found in the generated summary, while precision measures the percentage of n-grams in the generated summary that are also in the reference. The F1-score is the moyenne harmonique de précision et de rappel, fournissant une métrique unique qui équilibre les deux.
Overall, ROUGE Score is a valuable tool in the field of NLP, helping researchers and practitioners objectively measure the effectiveness of their text generation systems by providing insights into how well they replicate modèles d’écriture humaine.