Automatic Text Summarization Based on Transportation Network and Word Mover's Distances Embeddings: A Toy Experiment

Judith-Agueda Roldán-Ahumada, Martha Lorena Avendaño-Garrido, Juan-Manuel Torres-Moreno

Abstract: This paper presents CafeTAL, a new Automatic Text Summarization algorithm by extraction. CafeTAL is based on a Transportation Network model combined with Word Mover’s Distance using Word Embeddings. The document employed in the reported experiments is a French language document composed of two different topics, including polysemous words. In order to get significant statistics, our evaluation protocol is based on the ROUGE metric using a vast number (245) of human judges. Our results are very encouraging.