Research On Event Ontology Similarity Calculation And Its Application

With the arrival of the era of big data and artificial intelligence,the number of natural language web data has grown dramatically,and its data types were mostly in the form of narrative texts.We needed to operate these web data(such as data integration,data fusion,data filtering),of which the similarity calculation was an essential part.Ontology is an important natural language processing method,but there are many deficiencies in using traditional ontology similarity calculation methods to calculate the similarity between events.For example,it was difficult for traditional ontology to describe the complete information of an event class,and it was difficult to describe the relationship of event classes.Therefore,an event ontology similarity calculation method of was needed.The event ontology similarity calculation of was not only the basis of the techniques of event ontology integration,ontology concatenation and ontology merging,but also was the precondition of event clustering,event recommendation,and event semantic retrieval.Therefore,event ontology similarity calculation was worth studying.Event ontology similarity calculation could not be separated from the event ontology structure.So used the emergency domain as an example to construct the corresponding event ontology.Event ontology mainly included event class name,event class elements,event class classification relationships,and event class non-classification relationships,and used this information to calculate event ontology similarity.Event ontology similarity was widely used.This paper used of news personalization recommendation and heterogeneous data integration to discuss the application of event ontology similarity.The main research work of this paper is as follows:Event ontology constructionAt present,there are many deficiencies in using ontologies to deal with event ontology.It simply considered the event class as a concept and does not consider the event as an organic whole(such as,site of occurrence,object of participation,time of occurrence,etc.).It was also difficult to reflect the complex relationships(such as,causality,following relationships,concurrency,etc.)between events.This article used the emergency domain as an example to construct the event ontology.Firstly,according to the national classification of emergency,established upper-level events.Secondly,obtaining non-hierarchical relationships between event classes and event classes from various knowledge acquisition methods,established lower-level event classes.Then,used Protégé to complete the event ontology modeling,and used word2 vec to expand the instance of the event class.Finally,used a terrorist attack event as an example to analyze.The results show that the event ontology can clearly describe the event class,the event class element,semantic relationship between the event classes and has strong extensibility.Event ontology Similarity calculationThere were many deficiencies in using the traditional ontology similarity calculation method to calculate event ontology similarity.For example,didn't consider the event class' s elements information and event class as a unified whole and didn't consider non-hierarchical structures among event classes.This paper present a comprehensive method for event ontology similarity calculation.The method was based on words similarity,collections similarity,and hierarchical structures similarity.Event ontology similarity was discussed from the event class name,event class element,event class hierarchy,and non-hierarchical structure.Finally,got event ontology similarity.Experiments show that this method is more accurate than the traditional ontology similarity calculation method,and has richer semantic information,which is more in line with people's recognition of events.Application of event ontology Similarity calculation in personalized news recommendationIn order to better solve the problems of cold start,sparse data,lack of semantics,and low recommendation accuracy in the traditional recommendation system,this paper introduced the event ontology similarity calculation into the recommendation system.Firstly,used news corpora to construct event ontology structure.Then,used news feature extraction construct user interest model.Calculated the similarity between news events through the event ontology structure,and conduct personalized news recommendation based on the similarity of user interests and the semantic neighbors of news events.Calculated the similarity between news events according to event ontology structure,calculated the user's interest similarity based on the user interest model,and found relevant news events according to the semantic neighbors of news events.Personalized news recommendations based on the above three aspects.Experiments show that this method has greatly improved the accuracy,recall,and F-value of traditional methods.Application of event ontology Similarity calculation in heterogeneous data integrationIn the era of big data,news texts and multimedia data were processed in different ways and stored in different ways.This made it difficult to share and interoperate this data.This paper introduced event ontology similarity into heterogeneous data integration.First,the heterogeneous data were mapped into local event ontology.Then,calculated event class similarity in the local event ontology,classified similar event classes as one event class to associate with the event class of the global event ontology.Users access to heterogeneous data sources only need to access the global event ontology,rather than the underlying data format.Therefore,it solves the problem of heterogeneous data integration and provides an effective solution for heterogeneous data sharing and interoperability.