ES-ESens: Detection of Event Sentences Based on Evaluation of the Explicitness and Significance of Information

摘要

Extracting event sentences with explicit and significant information is a basic work of semantic analysis based retrieval of text information. Current methods of event extraction normally lack evaluation of the quality of information embedded in a sentence, thus the extracted event sentences usually contain inexplicit or insignificant information that is unusable for real-world applications. In this paper, we introduce ES-ESens, a deep learning based methodology for detecting event sentences with high-quality information. To evaluate the explicitness and significance of the information embedded in a sentence, ES-ESens adopts Recurrent Neural Network with attention mechanism to perform deep semantic analysis on the context of the sentence and the article that explains the background of the event. Based on datasets of practical applications, experiments are presented to show the performance of our methodology.