Discipline Ontology Learning And Semantic Annotation For Scientific Resources

With the develop of the Internet and software-hardware technique,scientific resources move from print to digital,and scientific resources are large and growing.This environment changes the literature getting mode and paper reading mode of the scientists.Firstly,scientists always spend lots of time in retrieval,selecting,and reading scientific resources.With the development of Web 3.0 and artificial intelligence,scientific navigation and retrieval tools need to give the feedback more intelligence,and realize semantic retrieval.On the other hand,scientists need to get as more information in the limited time,they need to get the fragment of the knowledge instead of reading full articles,so scientists need to get the fragments of literature and skipping from them.Key points to solve the above problems are ontology construction and semantic annotation.Ontology has a perfect structure aiming at semantic web,it provides the semantic that machine can understand,and support the retrieval extension.Semantic annotation can realize the linkage between the knowledge,and support the strategic reading of the scientists.But the discipline ontologies are in short,and manually ontology construction is time consuming and grueling.So,this paper use the ontology learning technology to construct the discipline ontology.Researches of ontology learning are focus on some special areas such as biomedical area or some special phrases(such as term extraction)of ontology learning,and the ontology learning methods for academic domain are shortage.On the other hand,the unstructured academic resources bring difficulties for the computer to retrieve.With the semantic annotation,computers can automatically understand the content and reasoning.But the annotation theory needs to be expanded.In order to solve the problems above,this paper mainly focus on the following works:Firstly,with the researching of the ontology learning,this paper proposes an ontology learning methodology for Chinese academic domain.This paper proposes the solutions for the four phrases of the ontology learning,that is,term extraction,concept formation,taxonomic learning and non-taxonomic learning.Combining linguistics,statistics and machine learning in ontology learning.In term extraction phrase,this paper proposes the term border recognition,obtains the rule of the structure of the Chinese academic terms,generates the statistical term sorting model,and transform the term sorting problem to predict problem by using Kriging and SVR.In concept forming phrase,this paper defines the types of the terms and constructing the taxonomy to form the concept.In relation learning phrase,this paper defines the types of taxonomic relationship and non-taxonomic relationship,combining template and graph method to obtain the taxonomic relationship,combining statistical method to obtain the concepts and triples.Finally,this paper leads the experiments to compare the methods proposed in this paper with existing methods,results show that the methods proposed in this paper have advantages among other methods.Secondly,this paper constructs the metadata for academic resources in type,content,structure and citation.This paper proposes the annotation methodology for every aspect of the academic resources.The metadata contains basic metadata,content metadata and citation metadata,and subdivide those metadata.Then annotate the content of the academic resources with combining manually annotation and automatic annotation.This paper proposes the ontology modularization to improve the effect of manually annotation,and annotate the content and citation with supervised methods.After all,this paper researches citation sentiment analysis.To improve the effect of supervised learning,this paper constructs several taxonomies.The methods are validated by the experiments.Works in this paper have a great value in academic domain ontology learning and semantic annotations,it can be used to other domains.First,this paper proposes a methodology for Chinese ontology learning.Second,propose a scheme for semantic annotation in academic resources.Subdivide the task of semantic annotation,and propose several methods.Finally,the academic domain ontology is the basic of the semantic retrieval.This paper is only focus on the concept learning and relation learning phrases.There is no farther study of axiom learning.The sematic annotation of semantic resources still has lots of tasks.In the further works,we need to solve the problems of axiom learning,and propose more solutions for semantic annotation.