Font Size: a A A

The Research Of Event Instance Inference Based On Extended Description Logic

Posted on:2015-12-16Degree:MasterType:Thesis
Country:ChinaCandidate:Y Y TangFull Text:PDF
GTID:2298330422489401Subject:Computer applications
Abstract/Summary:PDF Full Text Request
Event is an important information unit on the Internet. Compared withthe traditional concept, event is dynamic, and its status will change overtime. What’s more, there are some inherent relationships between events. Takingthe traditional concept as the unit of information representation on the Internetmay lead to Tennis Problems. As the unit of knowledge representation,event can describe more complex semantical information, and it isadvantageous to simulate the human brain’s working mechanism. Eventinference is multi-layed. Event instance checking is a fundamental and key issuein the event inferences. The other inferences need the support of event instancechecking. And some inferences can be transformed into event instance checkingin some ways. Description Logic is one of the most widely used methods ofknowledge representation, which receives a lot of attention for its benefits, suchas explicit model-theoretic semantics, effective concept hierarchy, decidableinference services and so on. However, the traditional description logic is onlyapplicable for static concept. When processing event knowledge, it has obviouslimitations. This dissertation analyzes the results of academic research on eventand summarizes the objective laws that human learn events and the laws ofevent performance in the text. Then combined with the previous work of theresearch group, it introduces the method of event representation based onextended describe logic and the method of event instance checking. The mainworks and contributions of this dissertation are as follows:(1) On the basis of the model of six-elements event, this paper introducesthe event knowledge representation based on an extended description logic,which not only could represent event class, but also could describe eventinstances. In the meantime, it supports to describe dynamic process of the event, the relationships between events and so on. And on the basis of this work, itgives the event instance status reasoning strategy based tense. Tense is animportant feature of event. When people learn events, they always concernwhether an event happens during a specified time. This dissertation concludesthe reasoning rules by event presentation features in the text, then it builds therelationships between time information and assertions through using these rulesand describes these assertions based on a temporal description logic. By this way,it can express the meaning that assertions of an event are available only during aspecified time. In this way, the event could express more semantic information.(2) Event instance checking is an important research task in event inference,which plays an important role in research of event ontology reasoning. Bystudying the features of event class and relations among event elements, thisdissertation proposes an event instance checking method based on extendeddescription logic. This method supplements semantic information at first, andthen it verifies whether an event is an instance of a specific event class by usingeffective reasoning capability of extended description logic, conceptsubsumption and satisfiability. An experiment is carried out to evaluate thisapproach. The experimental results show that this method can achieve instancechecking of event classes.(3) Research on Web service dynamic composition based on event ontologywhich is an application of the method of event instance checking. Most semanticdescription languages of Web service aim at individual services,and they rarelydescribe interactive relationships between services. A service not onlyinvolves the static information, but also includes some operations or entitieswhich could change the world, which is similar to the event in the real world.Event ontology is a formal, explicit specification of a shared event model thatexists objectively. In this dissertation, build mappings between Web service andevent class in event ontology and transform Web Service dynamic composition. Therefore, it could use the inference of event ontology to complete the goal ofWeb service dynamic composition.In conclusion, this dissertation proposes a method of event instancechecking which is used for the event in the text based on extended descriptionlogic depended on the objective laws of events. And this method is applied torealize Web service dynamic composition, so that it can effectively combine thetheory with the practice.
Keywords/Search Tags:Knowledge Representation, Event Ontology, Description Logic, Event Instance Checking
PDF Full Text Request
Related items