Research And Application Of Key Technologies In Text Event Relationship Extraction

"Event" is an independent process that takes place at a specific time and environment and is attended by several actors,exhibiting specific action characteristics,state change characteristics.It is an objective fact that reflects the interaction between people,things and things in the objective world.The occurrence of natural events is not often isolated,the occurrence and development are often associated with other natural events.These associations usually presented with the text description,often active in news reports,comments or Bowen and other media,some explicit relations,some hidden relationship.The semantics of the text is represented by the relationship between events and events,which is more in line with people's understanding of knowledge and contributes to the pre-judgment of the later development of the event.Through certain technical means,the machine can understand the semantic relationship between events.It is of great significance for emergency management decision-making,especially the handling of emergencies.The traditional relation extraction methods mostly utilize the shallow text features;focus on pattern matching and rules inferential,to extraction the limited semantic relations between entities.It is difficult to fit the large scale text relationship extraction;in the identification and determination of the relationship between events,the characteristics of the relatively narrow sense,the decision of the relationship category is more difficult.Therefore,the current problems in the research of text relations affect the accuracy and the degree of the relation extraction and recognition between the events,but also affect the judgment of specific logical relations between events.With the development of the relationship between changes in demand and application of Natural Language Processing technology,On the background of the study of the event of relation extraction was still not complete system and method,based on the traditional research of the event relation extraction,according to the text structure and content,deep semantic analysis research of the relationship between events is imperative.In this paper,event relation extraction and its application will be the main line.On the basis of traditional text representation model,event is proposed as a basic semantic knowledge unit to represent text.The characteristics of event distribution in text are explored.The event relation extraction and its phase in text information processing are studied by using relevant operational rules.Close the application.The main research contents can be summarized as follows:1)Text classification method based on Co-occurrence eventsAccording to the Chinese Emergency corpus(CEC),the paper deeply analyzes the distribution characteristics of the events in the text,explores the semantic knowledge of the relationship between the texts,use the phenomenon of co-occurrence between the various elements,with the algorithm to achieve the classification of text.2)Event relationship identification based on the Dependency and the Co-occurrenceThe relationship between the events and the events was the inherent attribute of the events,find out whether there was a logical relationship between events,which can help to predict the dynamic development of the event,especially the emergencies.Dividing the text used event representation,using the distribution characteristics of the event elements,the phenomenon of the co-occurrences overlap elements and the dependence relation between the text events,to construct a set of semantic event fine-grained clues collection,with the clustering algorithm to realization clustering of the relevant events,to achieve the semantic relations identification between events.3)Extraction causality relationships based on semantic eventA new method was proposed to identify the causal relationship in the text.The text is divided into events,extracting relevant events pair as a candidate events pair for determinant causal.According to the theory of causation,combining the causal event corpus performance model,explore the operation rules for calculating causal event class,determine whether there is a causal relationship between the event texts and determine the specific causal components.4)Automatic text summary representation model based on event semantic relation extractionOn the basis of event relation extraction,according to the relationship between events,the event network oriented graph was constructed to reflect the importance of the relationship between the events.The algorithm calculates the event importance degree of each node in the event network oriented graph and sorts it.With text events as the main line,according to the time sequence of events,export the event summary.The research work in this paper provides a new method for extracting and applying event relations in text,and plays a positive role in promoting event reasoning and related application research.