2015 Information Fusion_Faceted fusion of
RDF data organize notes
First, the paper organize your thoughts flow
1.1 Related research papers
- RDF distance matching fusion algorithm
- Fusion Algorithm Based on RDF inline
- Rule-based restrictions RDF fusion algorithm
- RDF data fragmentation: RDF data specific topics dispersed in a number of different data sets, each data set of RDF data that contains only one aspect of the subject matter.
- The dispersion of the RDF data integrated into the same subject in different ways according to
- RDF constructed from the results returned by search engines in TRG (specific topic RDF graph) and then use the RDF segmentation algorithm discovered a series of aspects.
- Using RDF segmentation algorithm discovered a series of aspects.
- Analysis for TRG: In the TRG, a node is o or s triad, while p is the corresponding triad, and both ends of one side vertices of a triplet. TRG is divided into two types of edges, and a reflection of the relationship between s o and the other reflects a relationship between two s. FIG TRG may be considered only between the edges of the structure is divided s FIG.
- On the data set used: a method of obtaining the field relating to the computer six data sets crawling data from the network, the first aspect of manual annotation each topic for later comparison
- FIG respect TRG obtained: found that the average degree of each node is the node 2 to 3.5,98% isolated; furthermore two vertices majority (96%) is connected to a point to the same aspect. Further Jaccard similarity calculated similarity score higher in accordance with the two nodes may point to the same node (thereby set a threshold value).
- Aspect of the discovery algorithm: RDF graph in FIG sub-divided into k disjoint, then the neighboring nodes according to similarity and homogeneity homogeneity node, TRG and is divided into data configuration diagram of FIG; data calculating a similarity according to FIG. series aspects referred ones; if in a side structure of FIG link two subgraphs, and two vertices belong to different aspects, it is possible to merge these two aspects.
- RDF discovery algorithm specific aspects of the process, see notebook
1.2 thesis problem
Papers problem-solving process 1.3
1.4 Experimental methods used paper
- About manual annotation: labeling rules given first, then two independent label with a theme, if the two conflict marked the emergence submitted to a third party marked and labeled as the final result.
- Conformity assessment: aspects for cell phones and manual annotation algorithm found set are compared using the contrast index NMI
- Evaluation found aspects: use of precision, recall, and Fl indicators evaluated before mentioned three methods are compared. Also evaluated between algorithms and manual annotation label
- FF proposed method than the previously mentioned three methods more effective (TRG considering the topological properties, and taking into account the degree of similarity and the topological properties)
- In F1, precision and recall these three indicators, six thematic aspects of extraction fusion, FF three methods are superior to the foregoing.
- The resolution and physical integration FF
- The FF integrate extended to other areas
The final evaluation of the results of experiments 1.5
1.6 follow-up paper
Second, the paper innovation
- Finally, some TRG Discovery Algorithm
- We found neighboring nodes and homogeneity homogeneity TRG similarity node topological properties of these two
- FF algorithm taking into account these two properties
Third, the proposed reference literature reading
[12]