利索-全球专利检索

  1. 登录
  2. 注册

搜索结果

3 条记录 (耗时 0.011 秒)

    METHODS AND SYSTEMS FOR INFERRING USER ATTRIBUTES IN A SOCIAL NETWORKING SYSTEM
    1.
    发明申请
    METHODS AND SYSTEMS FOR INFERRING USER ATTRIBUTES IN A SOCIAL NETWORKING SYSTEM 有权
    用于在社交网络系统中感染用户属性的方法和系统

    公开(公告)号：US20140279722A1

    公开(公告)日：2014-09-18

    申请号：US13839851

    申请日：2013-03-15

    申请人： Mitu Singh ,  Philip Bohannon ,  Jonathan Chang ,  Ismail Onur Filiz ,  Tudor Andrei Alexandrescu ,  Cameron Marlow

    发明人： Mitu Singh ,  Philip Bohannon ,  Jonathan Chang ,  Ismail Onur Filiz ,  Tudor Andrei Alexandrescu ,  Cameron Marlow

    IPC分类号： G06N5/04

    CPC分类号： G06F17/30702 ,  G06F17/30867 ,  G06N5/022 ,  G06N5/043

    摘要： A method and system for inferring user attributes in a social networking system. The method includes maintaining a social graph comprising a plurality of nodes and a plurality of edges between the nodes. An unknown, incomplete, or inaccurate user attribute for a user is identified, and a plurality of probability lists are generated using a corresponding plurality of probability algorithms that utilize known user attributes and the social graph. The probability lists include a set of probability entries, each including a prediction we value for the unknown, incomplete, or inaccurate user attribute and a confidence score. Using the probability lists and a plurality of weights corresponding to the probability algorithms, an inferred user attribute value is generated and stored. The weights may be adjusted based upon learning the correct value of the unknown, incomplete, or inaccurate user attribute, and search results may be modified to include the user for search queries seeking information about the inferred user attribute.

    摘要翻译： 一种在社交网络系统中推断用户属性的方法和系统。 该方法包括维护包括节点之间的多个节点和多个边缘的社交图。 识别用户的未知的，不完整的或不准确的用户属性，并且使用利用已知用户属性和社交图的对应的多个概率算法来生成多个概率列表。 概率列表包括一组概率条目，每个概率条目包括我们为未知的，不完整的或不准确的用户属性估计的预测值和置信度分数。 使用概率列表和对应于概率算法的多个权重，生成并存储推断的用户属性值。 可以基于学习未知的，不完整的或不准确的用户属性的正确值来调整权重，并且可以修改搜索结果以包括搜索查询的用户，寻找关于推断的用户属性的信息。