In order to fuse word-level and sentence-level information from different semantic spaces, the authors propose a hybrid optimization framework to optimize word-level information while simultaneously incorporate sentence-level information as constraints. The optimization is conducted by iterative unit substitutions. The performance on DUC benchmark datasets demonstrates the effectiveness of proposed framework in terms of ROUGE evaluation.
A qualitative method is presented for geographic information retrieval (GIR) to support qualitative representation, semantic matching, reasoning and ranking. The novel approach can avoid semantic information lost in current quantitative GIR methods. Information in documents and user queries are represented by propositional logic, which considers the thematic and geographic semantics synthetically. The similarity between documents and queries can be divided into thematic similarity and geographic similarity. The former is calculated by the weighted distance of proposition keywords in domain ontology, and the latter is further divided into conceptual similarity and location similarity which are measured by geo-ontology and spatial semantic respectively. Represented by propositions and information units, the similarity measurement takes evidence theory and fuzzy logic to obtain a general similarity from all sub similarities. This novel method retrieves qualitative geographic information from web and ranks documents semantically, which is consistent with commonsense, and thus can improve the efficiency of geographic information retrieval technology.
