Feature Analysis And Scenario Reasoning Method For Event Relation Detection

Event relation detection, as one of natural language processing technologies, faces information stream of texts detecting event relation. The event relation detection is aimed to identify and decide the relation type through analyzing the feature of semantic relevancy between events, which treats event as the basic semantic unit. The event relation detection includes event relation identification(identifying whether the event pair is related or not) and event relation type decision(deciding which relation between relevance events, e.g. cause relation). In the current study, it has not established a hierarchy of event relation detection research. Little research was carried through in the field of logic relation analyzing and processing. In this paper, we try to establish a system of event relation detection by learning from the concepts and data resources of discourse analysis, event extraction and scene understanding. Finally, we not only emphasize the analysis and comparison of the difference between event relation detection and discourse relation analysis, but also present the difficulty and challenge of the event relation detection. We also propose the methods of using event term and entity inference to recognize event relation and detecting event relation through cross-scenario inference.We presented the first study on these tasks, which presented as following:Firstly, event relation detection involves logic relation analyzing and processing of events. The study involved. In the current study, it has not established a hierarchy of event relation detection research. Aiming at this problem, we established a theoretical research system of event relation detection by learning from the concepts and data resources of discourse analysis, event extraction and scene understanding. The system includes the task definition, classification system of event types, corpora acquisition and annotations evaluation methodology, etc. We also focus on the difference between event relation detection and discourse relation analysis and give the difficulties and challenges of event relation detection.Secondly, event term and entity are important parts of events and can provide vital clues for event relation detection. In this section, we analyze the characteristics of the distribution of event term and event entity under the same topic. Our solution to event relation recognition is based on event term and event entity. The method resolved event relation into relation between event term and entity, thus enriching the line of reasoning. Our experimental results show that the method improved the performance for the tasks.Finally, event scenario is composed of ones which can describe an event. In this section, we propose cross-scenario inference for event relation detection. The core idea of our model is that mining the features of event pairs sharing same event scenario will help the clues construction for event relation detection. We first build scenario and relation for each event pair. And then the relation between events can be inferred by event scenarios.