Dependency Graph BasedSentence Fusion and CompressionVom Fachbereich Gesellschafts- und Geschichtswissenschaftender Technischen Universitaet Darmstadtzur Erlangung des Grades eines Doktors der Philosophie (Dr. phil.)genehmigte DissertationvonM.A. Ekaterina (Katja) Filippovaaus Sankt-Petersburg (Russland)Referent: Prof. Dr. Elke TeichKoreferent: Mirella Lapata (PhD, Reader)Einreichung: 25. Juni 2009Pru¨fung: 9. Oktober 2009D17Darmstadt2010iiAbstractThe popularity of text summarization (TS) in the NLP community has been steadilyincreasing in recent years. This is not surprising given its practical utility: e.g.,multi-document summarization systems would be of great use given the enormousamount of news published daily online. Although TS methods vary considerably,most of them share one important property: they are extractive, and the most com-mon extraction unit is the sentence – that is, most TS systems build summariesfrom extracted sentences. The extractive strategy has a well-recognized drawbackwhich is related to the fact that sentences pulled from different documents mayoverlap but also complement each other. As a consequence, extractive systems areoften unable to produce summaries which are complete and non-redundant at thesame time. Sentence fusion (Barzilay & McKeown, 2005) is a text-to-text gen-eration technique which addresses exactly this problem.