The use of MMR, diversity-based reranking for reordering documents and producing summaries
Proceedings of the 21st annual international ACM SIGIR conference on Research and development in information retrieval
The Earth Mover's Distance as a Metric for Image Retrieval
International Journal of Computer Vision
Improving Chinese tokenization with linguistic filters on statistical lexical acquisition
ANLC '94 Proceedings of the fourth conference on Applied natural language processing
Hi-index | 0.00 |
In this paper a new EMD-MMR (EMD: earth mover's distance; MMR: maximal marginal relevance) re-ranking method is proposed for generic Chinese text summarization. Our extraction-based summarization approach first ranks the sentences in a document by their weight calculated based on word frequency and position, and then re-ranks a few highly weighted sentences by the EMD-MMR method for sentence extraction. The proposed re-ranking method adopts a novel EMD-based similarity metric instead of the Cosine metric into the MMR approach. The EMD-based similarity metric can naturally take into account the semantic relatedness between words and compute the semantic similarity between texts with a many-to-many matching among words. We evaluate the performance of the proposed approach with a novel nk-blind method and the results demonstrate its effectiveness.