当前位置:首页 > 报告详情

URL 查找:回顾、进展和未来计划.pdf

上传人: Fl****zo 编号:718591 2025-06-22 10页 227.01KB

1、Hessisches Statistisches LandesamtURL finding:looking back,progress and plans for the futureWeb Intelligence Network Conference04.02.2025Heidi KhnemannStatistics Hesse2Hessisches Statistisches Landesamt URL finding in a nutshell What has been done(during the WIN):URL finding methodology reportWP 2 O

2、BEC annotation exercise Browser comparison(at Statistics Hesse)Plans and ideas for the futureOutline3Hessisches Statistisches LandesamtAutomated procedure to identify enterprise websites,usually containing these steps:sending search terms to a search engine,scraping the result URLs,extracting the re

3、levant information from the scraped data and applying a machine learning or rule-based model to link websites to enterprisesURL finding in a nutshell4Hessisches Statistisches LandesamtURL finding at the WIN:cooperation of WP 2 OBEC&WP 3 UC 5 Both WPs use enterprise websites as starting pointOBEC:cla

4、ssifiy online shops,social media presence,from enterprisewebsites for new indicatorsUC 5:classify economic activity codes and extract contact informationfrom enterprise websites to enhance the business registerCommon methodology report on URL findingSystematic collection of experiences and practical

5、 advice toimplement URL finding5Hessisches Statistisches LandesamtSome challenges with URL finding URL finding not possible on the WIH Scraping search engine result URLs is costly(e.g.storage,processingcapacities)URL finding is a difficult task,eg.because:o many to many relationship between enterpri

6、ses and websiteso Not all websites with enterprise data are the correct websiteso SBR data and website data are not always identical Despite all this:satisfactory accuracy of URL finding6Hessisches Statistisches LandesamtWP 2 OBEC:Manual creation of evaluation dataAnnotation exercise for URL finding

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
本文主要介绍了Hessisches Statistisches Landesamt(黑森州统计局)在Web Intelligence Network Conference上的报告,内容关于URL查找的进展、挑战和未来计划。关键点如下: 1. URL查找过程:通过搜索引擎发送关键词,抓取结果URL,提取相关信息,并使用机器学习或基于规则的方法将网站与企业关联。 2. 合作成果:WP 2 OBEC和WP 3 UC 5两个工作组共同制定了URL查找方法论报告,并分享了实践经验。 3. 挑战:URL查找在黑森州统计局内部不可行,抓取搜索结果成本高,且存在企业网站的多对多关系和正确性验证问题。 4. 核心数据:手动标注与自动化查找的URL匹配正确率在82.7%-97.4%之间。 5. 未来计划:实现URL查找过程的完全自动化,使用嵌套命名实体识别比较企业信息,定期检查和更新URL。 6. 其他议题:探索大型语言模型(LLMs)和开放网络搜索数据/欧洲搜索引擎的潜力。 参考文献和联系方式已在文末提供。
"如何自动寻找企业网站?" 哪个更适合数据抓取?" "未来统计办公室有哪些计划?"
客服
商务合作
小程序
服务号
折叠