Pattern-Based Knowledge Graph Embedding for Non-functional Requirements

Under the big data environment, knowledge of software non-functional requirements is connected with and distributed in large-scale, multi-sources, heterogeneous, multimodels and continuously growing data sources. In order to organize the information and establish a relationship between concepts so as to make knowledge available, we focus on building a knowledge graph of non-functional requirements. We found that the non-functional requirement knowledge is valid in specific fields. Considering the fact that traditional knowledge embedding model is not pleasant when dealing with one-to-many and manyto- one complex relationships in non-functional knowledge graphs, we will classify the domain of knowledge in non-functional knowledge graph by introducing patterns. Moreover, we are going to introduce pattern information into traditional knowledge embedding model, and propose a pattern-based non-functional knowledge embedding model-PNFE. The experimental results imply that PNFE model is superior to other traditional models in dealing with link prediction of non-functional knowledge graphs and quadruple classification tasks.