... 3) in a summarization application that aims at producingboth informative and fluent headlines. Our head-lines are generated in an abstractive, bottom-upmanner, starting from words and phrases. ... in a first phase,symbolic knowledge to (over)generate a large setof candidate realizations, and, in a second phase,statistical knowledge about the target language (such as stochastic language ... translations corre-sponding to a Chinese span , we use a prob-abilistic beam and a histogram beam to beamout low probability translation alternatives. At thispoint, eachspan is “tiled” with...