An Event Ontology Formalization Method Based On Elements Projection And Its Application In Sentiment Analysis

Regarding an event as a knowledge unit, knowledge representation and reasoning can be completed from internet text information, which is more close to the cognitive process of human beings. Existing methods of event knowledge representation and reasoning are mainly based on 5-W model, including who, what, when, where, how. However, there are some shortcomings in these methods as follows. Representation insulation of event elements results in weak semantic links between elements. Neglect of implicit element links between events reduces the ability of event knowledge inference. And it is difficult to unify event, element, status into a universal formal framework. To solve these problems, this paper proposes an event element projection method based on description logic SROIQ. This method attempts to transfer elements in an event to a complex concept of a specific element, converting event element reasoning into inference on concepts classes and roles relations in description logic. Based on it, event element projection is applied to formaliztion of event ontology pattern, to improve the contruction, representation and inference on domain event ontology. At last, this paper attempts to apply event element projection, as a brand new method, to sentiment analysis on internet news. The main work of this paper includes three parts as follows.(1) Representation and reasoning method of event knowledge is established. Due to the lack of effective methods of representation and reasoning for the event elements in text, and the deficiencies of traditional description logic expression on event knowledge, an event element projection is proposed as a new formalization method based on description logic SROIQ. This method attempts to represent and reason event, element, status in universal formal framework, and mine the element links and theorems to improve expression and reasoning ability in formalization framework. At same time, formalization method and its application example are provided based on element projection.(2) Event ontology pattern and its formalization method based on event element projection are proposed in this paper. Event ontology pattern is upper event ontology extracted from domain event ontologies, which contains common features in related domains. Based on previous work, this paper proposes type inherit, pattern inherit and event relation pattern, making representation and reasoning on event ontology pattern convenient and efficient. At same time, event ontology pattern provides a logical basis for sentiment analysis of news events.(3) An method of sentiment analysis based on event model, which predicts sentiment types of news readers through events in news texts. In this paper, it is called “events predicts sentiments”. Existing sentiment analysis of depends on comments on news, thus it is difficult to predict timely enough news readers’ sentiment. And traditional concepts and topic models fails to classify news associated with different sentiment types. This paper attempts to apply event element projection method to sentiment analysis, and proposes a sentiment analysis method based on event model in internet news. This method can predict public proficient sentiment expressed on specific via event classes, element classes, event relation pattern, etc.In summary, the event element projection method, proposed in this paper, can represent and reason on event knowledge based on description logic SROIQ, and be applied to sentiment analysis of news readers. This sentiment analysis method based on event model reduce dependence on comments texts, to improve the timeliness of public opinion monitoring, explore the semantic link between objective events and subjective feelings, and avoid negative influence resulting from irony texts.