Combining Text Embedding and Knowledge Graph Embedding Techniques for Academic Search Engines

Gengchen Mai, Krzysztof Janowicz, Bo Yan
2018 International Semantic Web Conference  
The past decades have witnessed a rapid increase in the global scientific output as measured by published papers. Exploring a scientific field and searching for relevant papers and authors seems like a needle-in-a-haystack problem. Although many academic search engines have been developed to accelerate this retrieval process, most of them rely on content-based methods and feature engineering. In this work, we present an entity retrieval prototype system on top of IOS Press LD Connect which
more » ... zes both textual and structure information. Paragraph vector and knowledge graph embedding are used to embed papers and entities into low dimensional hidden space. Next, the semantic similarity between papers and entities can be measured based on the learned embedding models. Two benchmark datasets have been collected from Semantic Scholar and DBLP to evaluate the performance of our entity retrieval models. Results show that paragraph vectors are effective at capturing the similarity and relatedness among papers and knowledge graph embedding models can preserve the inherent structure of the original knowledge graph and hence assist in link prediction tasks such as co-author inference.
dblp:conf/semweb/MaiJY18 fatcat:o2ehtvu7zjb4ljjraorwnxcnlq