Automated Trademark Monitoring Systems and IP Protection

商标为所有者提供了宝贵的知识产权。在世界范围内，侵权是一个日益严重的问题，网上有很多此类活动。在Ebay、阿里巴巴（Alibaba）、淘宝（Taobao）等网站上销售商标和假冒商品是众所周知的，而且一直是欧盟针对卖家和网站运营商采取法律行动的主题（参见蒂芙尼诉Ebay案）。2013年，英国知识产权局举办了一场名为“新技术如何减少……”的竞赛。侵权产品在网上的可及性显著提高。我们有一个全自动“突破性”系统的概念验证（POC），该系统可以通过任何销售清单中所有商标的相似性来监控侵权行为，每个商标的成本极低。使用机器学习来识别销售列表中的产品/品牌名称，它使用我们的美国专利商标相似度软件（参见我们的网站www.trademarkdirect.co.uk）将这些产品/品牌名称与所有相关商标的听觉和视觉相似性进行比较。这种比较比其他技术（例如Levenstein）要复杂得多，并产生具有更细粒度结果的距离测量，以自动建立侵权程度。拟议的项目将基于poc开发一个大容量、可扩展的原型。将为第二项功能开发额外的新算法，以区分合法转售的同品牌商品和假冒商品。这项开发工作需要原型将产生的大型数据集，但poc不能。

Trade marks offer valuable Intellectual Property rights to owners. Infringement is anincreasing problem worldwide with much such activity online. The sale of trade markinfringing and counterfeit items on websites such as Ebay, Alibaba, Taobao, is well knownand has been the subject of legal action both against sellers and website operators in the EU(see Tiffany v Ebay). The UK's Intellectual Property Office ran a 2013 competition 'How cannew technology..(reduce)..markedly the accessibility of infringing products online'.We have a proof of concept (POC) for a fully automated 'break-through' system that canmonitor for infringement through similarity of all trade marks by any sales listing for aminuscule cost per mark. Using machine learning to identify product / brand names in saleslistings it compares these for aural and visual similarity with all relevant trade marks usingour US patented trade mark similarity software (see our site www.trademarkdirect.co.uk).This comparison is considerably more sophisticated than other techniques (e.g. Levenstein)and produces distance measures with much more finely grained results to automaticallyestablish the degree of infringement.The proposed project will see development of a high volume, scaleable prototype based on thePOC. Additional new algorithms will be developed for a second function that distinguishesbetween identically branded items which are legitimate resales and those that are counterfeit.This development work requires the large data sets the prototype will produce, but the POCcannot.