Inter-Domain Linking of Problems in Science and Technology through a Bibliometric Approach

Abstract

Science and technology activities are recognized as problem-solving activities. Most solutions are created by tackling problems with previous knowledge, not only in an academic context but also in an industrial context. Scientific papers and patent publications can be regarded as explicit knowledge obtained by problem solving in the academia and industry, respectively. However, approaches toward problem solving do not necessarily match between scientific papers and patentable technology, even in the same field. The research question is addressed here is whether scientific problems can be provided insights from technical problems and solutions. In this study, we propose a concept to link problems in inter-domains for knowledge discovery using a linguistic approach. We extracted scientific papers and patent publications related to computer science as datasets in this study. Then, from these datasets, we identified problem sentences and solution sentences by neural probabilistic language model focusing on attention mechanism. Our approach is applied to extract groups of sentences for identifying semantically similar problems in inter-domains. From the results, we extracted several pairs of problem sentences across the domain. The results suggest that scientific problems and industry solutions may be able to give insights each other. This approach is also recommended not only for corporate activities but also for identifying research trends.