Classifying scientific publications according to Field-of-Science (FoS)
taxonomies is of crucial importance, allowing funders, publishers, scholars,
companies and other stakeholders to organize scientific literature more
effectively. Most existing works address classification either at venue level
or solely based on the textual content of a research publication. We present
SciNoBo, a novel classification system of publications to predefined FoS
taxonomies, leveraging the structural properties of a publication and its
citations and references organised in a multilayer network. In contrast to
other works, our system supports assignments of publications to multiple fields
by considering their multidisciplinarity potential. By unifying publications
and venues under a common multilayer network structure made up of citing and
publishing relationships, classifications at the venue-level can be augmented
with publication-level classifications. We evaluate SciNoBo on a publications'
dataset extracted from Microsoft Academic Graph and we perform a comparative
analysis against a state-of-the-art neural-network baseline. The results reveal
that our proposed system is capable of producing high-quality classifications
of publications.

本研究提出了一种新的科学文献分类系统 SciNoBo，它采用多层网络结构，基于文献结构和引文引用，将文献分类到预定义的 Field-of-Science（FoS）分类法中，并能够支持多个领域的多学科学术文献分类，并在来自 Microsoft Academic Graph 的数据集上表现出高质量的分类能力。

SciNoBo：一种基于层级多标签分类的科学出版物分类器

SciNoBo : A Hierarchical Multi-Label Classifier of Scientific Publications

We introduce a purely feed-forward architecture for semantic segmentation. We
map small image elements (superpixels) to rich feature representations
extracted from a sequence of nested regions of increasing extent. These regions
are obtained by "zooming out" from the superpixel all the way to scene-level
resolution. This approach exploits statistical structure in the image and in
the label space without setting up explicit structured prediction mechanisms,
and thus avoids complex and expensive inference. Instead superpixels are
classified by a feedforward multilayer network. Our architecture achieves new
state of the art performance in semantic segmentation, obtaining 64.4% average
accuracy on the PASCAL VOC 2012 test set.

引入了一种纯前馈体系结构用于语义分割，通过从超像素缩小到场景级别分辨率的方式，从嵌套的区域序列中提取丰富特征表示来映射小图像元素，并使用前馈多层网络对超像素进行分类，该架构在 PASCAL VOC 2012 测试集上取得了 64.4% 的平均准确率，是目前的最佳表现。