A Method For Obtaining And Analyzing Cross-language News Event Information In Non-ferrous Industries

With the continuous advance the strategy of 'One Belt,One Road',the internationalization of China's non-ferrous metals industry deepen further,Non-Ferrous Metal Enterprise pay more and more attention to events that are related to Product,enterprise,market and technology of Non-Ferrous Metals of different countries.Timely and comprehensive accessing these events and the relationship among them is essential for the non-ferrous industry,enterprise's risks,enhancing international competitiveness.Internet of different countries contains a bunch of news that cover these events.Non-Ferrous Metal Enterprise can know about relevant events and events correlation in time by these news.Using non-ferrous metals industry-related news in different languages of different countries on the Internet,we aims to automatic acquire related event information in different countries and analyze correlation between these events by computers.In the end,we can show the analysis results to users.Our research has an important role in the non-ferrous metals industry,enterprise business decision.Automatic accessing and analyze non-ferrous metal industry news event information on Internet,the key issue lies in:How to accurately identify non-ferrous metals industry related news in different languages.How to automatic integrate the relevant news of non-ferrous metals industry in different languages that cover the same events.How to automatically extract important event information from news events in different languages and get event summarization.How to automatically extract event information from different news reports to form a summary.How to analyze correlations among different events and show the analysis results.Due to the lack of bilingual resources in the nonferrous metals industry and the available machine translation tools,it is difficult for us to use the existing method.Besides,the existing methods do not take the characteristics of non-ferrous metals industry-related news into account,it is difficult to achieve good results.So we research cross-lingual news clustering;cross-lingual news digest;news event correlation identification;relational graph construction and achieve the following innovative results.(1)Proposed a method based on CNN.Which can identify cross-lingual Nonferrous Metals Industry News.For the problem that the existing cross-lingual news identification method didn't consider the domain features,we use bi-lingual dictionary to build cross-lingual word vector to represent different language news.On this basis,using CNN to build a model which can extract and identify features of different language Nonferrous Metals Industry News.Experiment results show that this method can effectively increases Identification Effect of Nonferrous Metals Industry News recognition.(2)Proposed an event-elements based method that can cluster different language Nonferrous Metals Industry News.For the problem that existing cross-lingual news clustering method didn't consider news events information.We use non-ferrous metal enterprises,products and other information in non-ferrous metal industry news as event element to characterize the news.Evaluate the similarity of different language news by comparing the event elements contained in different language news.And then on this basis,using incremental clustering method to cluster online incremental news.Experiments show that this method can effectively improve the automatic classification effect of nonferrous metals industry news.(3)Proposed a cross-lingual news summarization method combining nonferrous metals industry bilingual topic model and graph model.For the problem that existing methods rely on machine translation tools and don't use both news topic information and the relationship between the sentences.We use the translation domain terms as a cross-lingual bridge to mine bilingual topics of different language news.Use cross-lingual word vector to build sentence-related graphs to rate the importance of sentences under different topics.At the same time,choose the summarization sentence according to the semantic similarity and the importance of the sentences under different topics.The experimental results show that the proposed method can automatically summarize the non-ferrous metals industry news in different languages and enhance the effect of automatic summary.(4)Proposed explicit and implicit cross-lingual event correlation identification method.On this basis we propose and construct event association graph which fuse time information.This graph can represent before-after correlation end implicit-explicit correlation of events.For the problem that existing event correlation identification method apply in the monolingual environment and apply for word or sentence correlation identification.Use Different language news that covers events as a basis for judging whether events are relevant.The explicit method recognize event correlations based on characters appears in different related news.The implicit method is based on similarity of event elements.The experimental results show that the method has achieved good effect for identifying correlation among cross-lingual news events.Finally,build an event related graph based on relevant events identification that fuse time information to characterize the association among all events.