Event Extraction from News Media Based on Knowledge and Data-Driven Methods

The paper analyzes up-to-date methods for Event Extraction from unstructured data. The proposed approach is based on both knowledge and data-driven methods for Event Extraction, in particular, from news data streams displaying the agenda. Firstly, syntactic dependency relations between extracted components of events are identified, representing coherence of the text fragments (collocation candidates). T-score association measure is applied at the next stage of collocation extraction. The advantage of the developed technology is that both the grammatical structure of the language and the statistical association between event components are analyzed. The effectiveness of the developed technology is assessed and exceeds the precision results of the other approaches. The benefits of the proposed technology are to improve the quality of news stream processing.