A Research Of Multi-function Switched-beam Antenna Array

With the development of the information age,people's dependence on the Internet is increasing.More and more people express their opinions or record their lives on social network.Therefore,the analysis and research based on social network are becoming more and more active.Event detection based on social network can discover events happening in real life at the first time,which is conducive to enhancing the capabilities of sudden warning and public opinion monitoring.The event element extraction analyzes the event text and extracts information about the person,place,time and other information related to the event,so that people can understand the event more intuitively and master the core information in the event.The existing event information extraction can be divided into two methods: pattern matching and statistical learning.The pattern matching method matches the text content by formulating rules,usually applicable to specific fields,and the rule formulation cost is high,so in practice it is difficult to apply on a large scale;the statistical learning method extracts information such as person name,place name,time,etc.in the event by identifying the name entity in the event text,but the existing name entity recognition method generally has the problem that the recognition type is too large,resulting in low accuracy.And the recognition effect of the social network short text in the noisy environment is not ideal,so it is difficult to obtain accurate event element information.This thesis studies the method of extracting event elements in social network,focusing on the extraction of time elements and geographic elements.The main contributions and innovations are as follows:(1)A time element extraction method based on model constraints is proposed.This method overcomes the problem that the traditional name entity recognition method extracts features are not targeted,and adds feature sets according to the characteristics of time elements in the social network.At the same time,in the process of using the conditional random field model to identify time elements,in view of the learning speed is too slow and inaccurate boundary identification problem,conditional random field model with constraints is constructed.After experimental data testing,this method can improve the accuracy of identifying time elements,thus this method improves the overall performance of the model.(2)A geographic location elements extraction method based on text syntax features is proposed.Firstly,the method constructs the consistency model between the geographical name entity and the geographic location elements of the event,and extracts the candidate set from the tweet collection of the event.Secondly,in the extraction process,in order to solve the problem of sample imbalance,the identification accuracy is improved by weighting the samples.Finally,this thesis takes the consistency probability as the weight,extracts the location in the candidate set as the event geographical location element,and then maps the extracted geographical name entity into GPS for comparison,which demonstrates the accuracy of the method.