Affiliations: [a] Department of Information Science, Faculty of Humanities and Social Sciences, Khon Kaen University, Khon Kaen, Thailand | [b] Informatics Innovative Center of Excellence, Walailak University, Thai Buri, Nakhon Si Thammarat, Thailand | [c] Rajamangala University of Technology Srivijaya, Thung Song, Nakhon Si Thammarat, Thailand
Abstract: This article describes the development process of the Thai cultural knowledge graph, which facilitates a more precise and rapid comprehension of the culture and customs of Thailand. The construction process is as follows: First, data collection technologies and techniques were used to obtain text data from the Wikipedia encyclopedia about cultural traditions in Thailand. Second, entity recognition and relationship extraction were performed on the structured text set. A natural language processing (NLP) technique was used to characterize and extract better textual resources from Wikipedia to support a deeper understanding of user-generated content by using automatic tools. Regarding entity recognition, a BiLSTM model was used to extract relationships between entities. After the entities and their relationships were obtained, triple data were generated from the semistructured data in the existing knowledge base. Then, a knowledge graph was created, knowledge bases were stored in the Neo4j Desktop, and the quality and performance of the created knowledge graph were assessed. According to the experimental findings, the precision value is 84.73%, the recall value is 82.26%, and the F1-score value is 83.47%; therefore, BiLSTM-CNN-CRF can successfully extract entities from the structured text.
