Scientific publications

Тарасов С.Д.

Исследование и оптимизация параметров алгоритма Manifold Ranking на основе метрики автоматической оценки качества обзорного реферирования ROUGE-RUS

// Электронные библиотеки: перспективные методы и технологии, электронные коллекции: Труды XI Всероссийской научной конференции RCDL'2009. Петрозаводск: КарНЦ РАН, 2009. C. 86-93

Tarasov S.D. The research and parameter's optimization of Manifold Ranking Algorithm based on automatically summarization evaluation metric by ROUGE-RUS // Digital Libraries: Advanced Methods and Technologies, Digital Collections: Proceedings of the XI All-Russian Research Conference RCDL'2009. Petrozavodsk: KRC RAS, 2009. Pp. 86-93

In this article we review the ROUGE metric adopted by DUC for automatic summarization evaluation and also it modification for Russian language ROUGE-RUS implemented by author.
Developed prototype with WEB interface for automated evaluation of multi-document summarization performance based on ROUGE-RUS
Concluded the experiment with objective to compose the handmade annotations for news clusters and automated evaluation of multi-document summarization.
Defined the base variable ROUGE-RUS on cluster. Reviewed the algorithm of Multi-document summarization "Manifold Ranking".
Researched the impact of diversity penalty (base parameters and initial input) on algorithm final result. The algorithm was modified to compensate above diversity penalty.

The research and parameter's optimization of Manifold Ranking Algorithm based on automatically summarization evaluation metric by ROUGE-RUS (611 Kb, total downloads: 123)

Last modified: October 16, 2009