Paper: Natural Language Generation and Summarization at RALI

ACL ID W13-2110
Title Natural Language Generation and Summarization at RALI
Venue European Workshop on Natural Language Generation
Year 2013

either abstractive or extractive, the former being the selection of the most important sentences from the original documents. In much the same way as NLG has suffered from the fact that it is often possible to trick the readers with canned text or formatted templates, abstractive summarization had to compete with ac- ceptable results produced by scorers of sentences, the ones with the best scores being then concate- nated to produce a summary. In our group, we tried to stay away from such approaches that in our view did not give any new insights even though it did not always allow us to win the summarization competitions at DUC or TAC. SUMUM (Saggion and Lapalme, 2002) ex- plored the idea of dynamic summarization by tak- ing a raw technical text as input and produced an indicative-informat...