Yuan Yin — Page 3 | Searcharxiv

arXiv2021

Patent-KG: Patent Knowledge Graph Use for Engineering Design

To facilitate knowledge reuse in engineering design, several dataset approaches have been proposed and applied by designers. This paper builds a patent-based knowledge graph, patent-KG, to represent the knowledge facts in patents for engineering design. The arising patent-KG approach proposes a new unsupervised mechanism to extract knowledge facts in a patent, by searching the attention graph in language models. This method avoids using expensive labelled data in supervised learning or listing complex syntactic rules in rule-based extraction. The extracted entities are compared with other benchmarks in the criteria of recall rate. The result reaches the highest 0.9 recall rate in the standard list of mechanical engineering related technical terms, which means the highest coverage of engineering words. The extracted relationships are also compared with other benchmarks. The result shows that our method provides more contextual information in relationships, and extracts more relationship types including positional and negation relationships.

cs.IR↗

arXiv2019

Scalable Graph Embeddings via Sparse Transpose Proximities

Graph embedding learns low-dimensional representations for nodes in a graph and effectively preserves the graph structure. Recently, a significant amount of progress has been made toward this emerging research area. However, there are several fundamental problems that remain open. First, existing methods fail to preserve the out-degree distributions on directed graphs. Second, many existing methods employ random walk based proximities and thus suffer from conflicting optimization goals on undirected graphs. Finally, existing factorization methods are unable to achieve scalability and non-linearity simultaneously. This paper presents an in-depth study on graph embedding techniques on both directed and undirected graphs. We analyze the fundamental reasons that lead to the distortion of out-degree distributions and to the conflicting optimization goals. We propose {\em transpose proximity}, a unified approach that solves both problems. Based on the concept of transpose proximity, we design \strap, a factorization based graph embedding algorithm that achieves scalability and non-linearity simultaneously. \strap makes use of the {\em backward push} algorithm to efficiently compute the sparse {\em Personalized PageRank (PPR)} as its transpose proximities. By imposing the sparsity constraint, we are able to apply non-linear operations to the proximity matrix and perform efficient matrix factorization to derive the embedding vectors. Finally, we present an extensive experimental study that evaluates the effectiveness of various graph embedding algorithms, and we show that \strap outperforms the state-of-the-art methods in terms of effectiveness and scalability.

cs.LG↗