A Study on Graph-based Classification for Important Technical Documents

In an era of technological hegemony, protecting innovative technologies worldwide has become an important issue. In particular, technical documents, including national core technologies, need to be protected separately from other general technical documents to maintain national technological competitiveness. Many studies on the classification of technical documents have been developed to analyze technological trends and classify the documents according to their types. However, since the existing methods have not considered the importance of documents, there is a limitation in that it is difficult to protect sensitive information such as national core technology. Therefore, in this study, we propose a new graph-based methodology that can perform binary classification of technical documents based on their importance. On top of that, we focus on explaining three steps of the proposed process using the properties of the graph network: graph representation, keyword extraction, and graph-based classification. We experimented with extracting keywords from each display and OLED document and classifying the test OLED document using PageRank and betweenness centrality. In conclusion, we show that our proposed method can effectively protect important documents in the technology industry by correctly classifying the class of a new document through the keywords we extracted from our experiments.