Research On The Abstracting Temporal Expression Based On The Terrorist Attack Events Database

During the 20th century,computers and the Internet has been rapid development,and information emerged on the exponential growth no matter in the quantity or speed of propagation.This development drives people to use the more efficient way to get information from the Internet,so more and more researchers focus on the work of natural language processing.Meanwhile,in the economic globalization today,security concerns of the international community have also been national attention.On the Internet,reports about terrorist attacks are also increasing.If researchers can properly use the results of natural language processing,and extracting information about this topic,in order to combat terrorism and make certain warning reference.Therefore,the paper chooses the subject of terrorist attacks,the specific research information extraction time extracting phrases.First,the paper gives a brief introduction of the research background,purpose and significance of the lead article.Increasing number of terrorist attacks,in order to give a hand to government agencies to provide relevant historical data,this article extracts the existing information about the terrorism topic.Analyzing the relationship between the time of each event to identify the perpetrators of terrorist organizations time habit and data also has a high predictive value.Certain training and testing procedures are also provided in this article,which extensively illustrates simple features and complex features and fully stating its whole structure and corresponding criteria,After concrete experiments,this article proves that the capacity of complicated templates are better than those simple ones,of which the complex features template's F was 85.74%and simple features template's F was 85.32%.Second,in the literature review chapter,article gives the detinution ot the time and introduces the English theory of time,elaborated on relevant research questions in English time identification.Meanwhile author analyzes the advantages and disadvantages of several methods of identification.Then the article introduces the research status of domestic decimation in time terms.In the end,it introduces some machine learning models used in the paper..Next,from all the data provided from the database,with those features and contents,certain plans to analyze news of terrorist attacks(from 2005 to 2014)were made.Based on those corpuses,certain news corpuses are built based on those information above.Using tools to set tags on those 100 news eorpuses as testing eorpus.A display platform is provided in this article,focusing on its function and further improvement problems.Finally,Certain training and testing procedures are also provided in this article,which extensively illustrates simple features and complex features and folly stating its whole structure and corresponding criteria.After concrete experiments,this article proves that the capacity of complicated templates are better than those simple ones,of which the complex features template's F was 85.74%and simple features template's F was 85.32%.Finally,author researches on the knowledge of extraction time phrase structure.The importance of temporal phrase is the inner frontiers and the function of sentence structure.Then using CRF and building up the initial characteristics and complexity of the characteristics templates.Adding these two platforms to the CRF model.Through specific experiments,the performance of complex template is better than the initial template.Wherein the complex template F is 85.74%,while initial template F value was 85.32%...