SBIR Phase I: Learning Consumer Preferences through Semantic Analysis for Hybrid Personalized Recommendation on the Social Web

小企业创新研究（SBIR）第一阶段项目解决了与可供消费的商品/服务数量激增相关的机会。这项拟议中的技术如果成功，将通过创新地识别用户偏好背后的原因，预测哪些商品/服务最能满足个人消费者，从而为消费者在可上网的智能手机上节省时间、精力和金钱。该方法利用语义分析来评估数以百万计的在线评论，确定为什么人们一般喜欢/不喜欢特定的娱乐机会，如餐馆、现场活动、俱乐部或表演。然后，它开始创建一个丰富而细致的用户档案，利用用户意见的语义分析，对先前的反馈及其特征的隐式或显式反馈，以及行为跟踪。随着技术的预测，包括用户自由文本在内的反馈机制不断提高对用户的理解，提高推荐的性能。预测是与上下文相关的，结合了行为和上下文信息，如GPS位置数据、天气状况、最近的搜索查询，以及从用户的社交网络档案和互动中挖掘的隐含的兴趣指示。这种方法旨在通过回答“你喜欢什么样的餐厅、酒吧、俱乐部或现场活动”这个问题，为用户提供“惊喜和愉悦的精彩预测”。对于广告商来说，它提供了前所未有的客户细分水平，只向感兴趣的个人投放广告。由广告商赞助的商品/服务的推荐是通过互联网和具有GPS功能的智能手机发送的。通过特殊的促销活动（如优惠券和折扣），移动应用程序跟踪消费者在广告商物理位置的购物行为。这为广告主生成了简单可靠的转换指标，跟踪实际销售情况并直接阐明广告成本效率。到目前为止，广告与实际客户销售/新客户引导之间的直接因果关系的证据主要局限于互联网，在那里可以跟踪横幅广告点击到在线购买。拟议的方法将创新的基于销售业绩的广告和在线效率带到现实世界的实体购买中，如果成功，将解决一个重要的和不断增长的机会。

The Small Business Innovation Research (SBIR) Phase I project addresses the opportunity associated with the explosion in the number of goods/services available for consumption. The proposed technology is one that, if successful, will predict which goods/services would optimally satisfy individual consumers by innovatively identifying the why behind users' preferences, to save time, effort and money for consumers on their web-enabled smartphones. The approach utilizes semantic analysis to evaluate millions of online reviews, identifying why populations in general like/dislike specific entertainment opportunities such as restaurants, live events, clubs, or shows. It then begins the creation of a rich and nuanced User profile utilizing semantic analysis of user opinions, implicit or explicit feedback on previous feedback and their characteristics, and behavioral tracking. As the technology makes predictions, feedback mechanisms including free text by the user continually improve understanding of the User and improve the performance of the recommendations. Predictions are made contextually relevant, incorporating behavioral and contextual information such as GPS location data, weather conditions, recent search queries, and implicit indications of interest mined from the user's social network profiles and interactions. The approach seeks to offer "brilliant predictions that surprise and delight" its users, by answering the question, "What restaurant, bar, club or live event would you love" For advertisers, it provides previously unseen levels of customer segmentation, advertising only to interested individuals. Recommendations of goods/services sponsored by the advertiser are delivered via internet and GPS enabled smart phones. Through special promotions (such as coupons and discounts), the mobile application tracks consumer shopping behavior at the physical location of the Advertiser. This generates simple robust conversion metrics for the advertiser, tracking actual sales and directly elucidating advertising cost efficiency. Heretofore, proof of direct causal linkage between advertising and an actual customer sale/new customer lead was primarily limited to the internet, where banner ad click through to an online purchase could be tracked. The proposed approach brings innovative sales performance based advertising and online efficiency to physical purchases in the real world and if successful will address a significant and growing opportunity.