Construction of Social Community Knowledge Graph from Wikipedia

Knowledge graphs have become a fundamental trend in the fields of healthcare, education, finance, ICT, science, engineering, society, politics, and tourism, etc. The community knowledge graph aims to explore the composition of communities and their relationships through visual social network analysis. However, the process of unstructured data like text raises challenges for knowledge extraction. To automatically build a knowledge graph, this study applies natural language processing and technology of knowledge graph creation. This paper proposes a novel method, Augmented Word Vectors (AWV), which is a word vector representation with lexical features and part-of-speech (POS) features. This work evaluates the performance of AWV on various deep learning models, and the experimental results show that AWV is effective for improving word embedding in the task of named entity recognition. Furthermore, an unsupervised approach based on rules and dependency semantics was applied to extract relations. Specifically, this method achieves high productivity by mapping sentences as dependency trees to extract relations mediated by verbs or nouns. We build a community knowledge graph for Hong Kong political groups by applying our method to the Chinese Wikipedia, which validates the effectiveness of the proposed method.