Purpose:This study aims to construct an ontology to model the semantics of social media streams,in particular,trending topics and public issues.Design/methodology/approach:Our knowledge base included 10 public events ...Purpose:This study aims to construct an ontology to model the semantics of social media streams,in particular,trending topics and public issues.Design/methodology/approach:Our knowledge base included 10 public events and topics from Weibo respectively,which were collected through keyword search and a crawler program.We used a semi-automatic approach to model and annotate the semantics in social media,and adapted the multi-layered ontology to refine the design based on previous researches,then we used named entity recognition(NER) to extract entities to instantiate the ontology.Relationships were extracted based on co-occurrence measures.Finally,we manually conducted post-filtering evaluation and edited the extracted entities and relationships.Findings:An initial assessment demonstrated that our multi-layered ontology supports various types of queries and analyses in the public issue knowledge base(PIKB),which can serve as an effective tool to query,understand and trace public issues.Research limitations:Manual involvement cannot meet the requirements for challenges of sustainable developments.Since the relationships extracted are fully based on the co-occurrence of entities,rich semantic relationships,such as how much the key players have been involved,could not be fully reflected.Besides,the user evaluation is necessary for further ontology assessment.Practical implications:The PIKB can be used by regular Web users and policy makers to query,understand,and make sense of public events and topics.The methodology and reusable ontology model are useful for institutions that are interested in making use of the social media data.Originality/value:In this study,a multi-layered ontology is applied to model the evolving semantics of public events and trending topics in social media,and the semi-automatic approach could make it possible to extract entities and relationships from large amount of unstructured short texts of user generated content(UGC) from social media.展开更多
基金supported by Beijing Thinker Workshop(Grant No.XK201211001)
文摘Purpose:This study aims to construct an ontology to model the semantics of social media streams,in particular,trending topics and public issues.Design/methodology/approach:Our knowledge base included 10 public events and topics from Weibo respectively,which were collected through keyword search and a crawler program.We used a semi-automatic approach to model and annotate the semantics in social media,and adapted the multi-layered ontology to refine the design based on previous researches,then we used named entity recognition(NER) to extract entities to instantiate the ontology.Relationships were extracted based on co-occurrence measures.Finally,we manually conducted post-filtering evaluation and edited the extracted entities and relationships.Findings:An initial assessment demonstrated that our multi-layered ontology supports various types of queries and analyses in the public issue knowledge base(PIKB),which can serve as an effective tool to query,understand and trace public issues.Research limitations:Manual involvement cannot meet the requirements for challenges of sustainable developments.Since the relationships extracted are fully based on the co-occurrence of entities,rich semantic relationships,such as how much the key players have been involved,could not be fully reflected.Besides,the user evaluation is necessary for further ontology assessment.Practical implications:The PIKB can be used by regular Web users and policy makers to query,understand,and make sense of public events and topics.The methodology and reusable ontology model are useful for institutions that are interested in making use of the social media data.Originality/value:In this study,a multi-layered ontology is applied to model the evolving semantics of public events and trending topics in social media,and the semi-automatic approach could make it possible to extract entities and relationships from large amount of unstructured short texts of user generated content(UGC) from social media.