In this research, we propose a particular version of KNN (K Nearest Neighbor) where the similarity between feature vectors is computed considering the similarity among attributes or features as well as one among values. The task of text summarization is viewed into the binary classification task where each paragraph or sentence is classified into the essence or non-essence, and in previous works, improved results are obtained by the proposed version in the text classification and clustering. In this research, we define the similarity which considers both attributes and attribute values, modify the KNN into the version based on the similarity, and use the modified version as the approach to the text summarization task. As the benefits from this research, we may expect the more compact representation of data items and the better performance. Therefore, the goal of this research is to implement the text summarization algorithm which represents data items more compactly and provides the more reliability.
Ultraviolet and x-ray-photoemission spectra (XPS) as well as x-ray inverse-photoemission [bremsstrahlung isochromat spectroscopy (BIS)] and x-ray-absorption spectra of valence-fluctuating are discussed to clarify its electronic structure. The constant initial state and constant final state spectra in the Sm 4d→4f excitation region are quantitatively interpreted in terms of the 44→44 and 44→44 multiplet structures. The Sm 3d absorption spectra support the bulk valence fluctuation. The 4 and 4 multiplet structures are clearly resolved in the 4d XPS spectra. The effective hybridization of the Sm 4f states with the conduction-electron states is found to be large and much different for the 3d photoemission, 4d photoemission, and 4f BIS final states, resulting in the different intensity ratio of the and signals in these spectra. In addition, the surface core-level shift of the Sm 4 final state is evaluated as ∼0.5 eV.