Knowledge Scanning And Application Discovery Of Knowledge Atlas

Knowledge graph and knowledge base have been widely used in various kinds of applications, such as semantic web, machine translation, query expansion, text classification and so on. In addition, taxonomies and ontologies have been playing an important role in data cleaning, entity resolution and other fields of information integration. As the scale of knowledge graphs and knowledge bases grow, on one hand, there might be data quality problems. On the other hand, how to store such large scale data and how to find more application for knowledge graph and knowledge base is a great challenge for research in knowledge graph.In this paper, we first focus on the "orphan concept" problem in knowledge graphs and then make a certain degree of exploration about how to store knowledge and how to use knowledge bases in software engineering.For the challenge to data cleaning problem in large scale knowledge graph, we first find that there exist "orphan concept" in knowledge bases like ProBase and FreeBase by case study. Then we use statistical analysis based algorithm to find all these "orphan concepts". By natural language method and machine learning method, we successfully add missing links between "orphan concepts" and their hypernyms. Experimental results show that our results can improve knowledge graph accuracy. Then we make exploration about how to store and apply knowledge graphs. Based on the idea of re-ordering indices, we successfully compress a biology knowledge base. And in the XML attribute value recommendation scenario, we by using mining techniques and semantic rank based the results of word breaking, we develop a novel method to recommend developer correct attribute values. Evaluation results show that our tool can actually propose good quality XML attribute values, which significantly improves their efficiency in building framework based applications.