Introduction: Knowledge Graph Building
Information is being produced at speed well beyond human consumption in contemporary society. Knowledge graphs have become an important representing way of specific domain information. However, data volume, distributions and inner linkages of Specific-Domain Knowledge Graph(SKG) and Open-Domain Knowledge Graph(OKG) are totally different. Therefore, human interferences play a crucial role in SKG building. Based on unstructured text, semi-structured data and knowledge bases of third party, we aim to build a knowledge graph by knowledge acquisition, knowledge fusion, knowledge graph completion, knowledge graph demonstration and publish technologies. It is conductive to realize knowledge fusion, afford research proposes for professionals, and furthermore promote developments of corresponding areas.
[Top]        
Motivation
Compared with more matured building technologies and process of OKG, SKG relies a lot on human, getting more accurate results but lower efficiency. Therefore, our system aims to develop SKGs. Main difference of OKG and DKG is data distribution having a decisive effect in knowledge graph building. By improving traditional models and methods, we get great performance in SKG building.
[Top]        
Research work
  • Entities Disambiguation
  • An entity conceptualize method combing topic based models and knowledge bases : Given a query, we first use a dictionary constructed from the knowledge bases to detect the possible entities and their associated categories. Then, we use a topic based method to derive semantic information from the text. By comparing the topical similarity between various candidate phrases, we get the most likely entities and their related categories.

    An entity linking method based on Markov Logic Network: This Method links the entity mentions in scientific text to appropriate concepts in the domain ontology by Markov Logic Network. Combining the first-order logic and probability model, the framework enables us to integrate many domain specific features such as the domain naming rules as well as common context features and entity coreference features, thus provides an effective way to deal with the challenges of the name diversification, abbreviation and other complicated naming rules of entities in specific domains.

    [Top]        
    Grant
    • 2016.01-2020.12 Research on Key Technologies of Large-Scale Linked Data Management Granted by the Natural Science Foundation of China(NSFC) under grant number 61532010
    • 2016.07-2019.06 Large-scale Knowledge Graph Research on Microorganism Area Granted by the National Key Research and Development Program of China ‘Scientific Big Data Management System’ under grant number 2016YFB1000603
    [Top]        
    Publication
    • Z. Du, Z. Hao, X. Meng, et al. CirE: Circular Embeddings of Knowledge Graphs[C]. Proceedings of the International Conference on Database Systems for Advanced Applications (DASFFA), pages:148-162, 2017, Suzhou, China.
    • Z. Hao, Z. Wang, X. Meng, et al. Semantic Definition Ranking[C]. Proceedings of the International Conference on Database Systems for Advanced Applications (DASFFA), pages:153-168, 2017, Suzhou, China.
    • K. Zhao, X. Meng, H. Li, Z. Wang: Using Encyclopedic Knowledge to Understand Queries[C]. Proceedings of the 24th ACM International Conference on Information and Knowledge Management (CIKM), pages:17-22, 2015, Melbourne, Australia.
    • Z. Wang, K. Zhao, H. Wang, X. Meng, J. Wen: Query Understanding through Knowledge-Based Conceptualization[C]. Proceedings of the 24th International Joint Conference on Artificial Intelligence (IJCAI),pages:3264-3270, 2015, Buenos Aires, Argentina.
    • 孟小峰. 大数据管理概论[M]. 北京:机械工业出版社, 2017.
    • (美)Christine L. Borgman 著. 大数据、小数据、无数据[M],孟小峰,张祎,赵尔平,译,北京:机械工业出版社,2017.
    • Xin Luna Dong(董欣) 著. 大数据集成[M],王秋月,杜治娟,王硕,译,北京:机械工业出版社,2017.
    • (美)AnHai Doan, Alon Halevy, Zachary Ive著. 数据集成原理[M],孟小峰,马如霞,马友忠等,译,北京:机械工业出版社,2014.
    [Top]        
    Maintained by Zhongyuan Wang() Copyright © 2007-2009 WAMDM, All rights reserved