公开(公告)号：DE60044220D1

公开(公告)日：2010-06-02

申请号：DE60044220

申请日：2000-01-24

Applicant: XEROX CORP

Inventor： CHEN FRANCINE R ,  SCHUETZE HINRICH ,  GARGI ULLAS ,  PITKOW JAMES E ,  PIROLLI PETER L ,  CHI ED H ,  LI JUN ,  NILES LESLIE T

IPC: G06F17/30 ,  G06F12/00

公开(公告)号：JP2010262662A

公开(公告)日：2010-11-18

申请号：JP2010140220

申请日：2010-06-21

Applicant: Xerox Corp ,  ゼロックス コーポレイションXerox Corporation

Inventor： CHEN FRANCINE R ,  SCHUETZE HINRICH ,  GARGI ULLAS ,  PITKOW JAMES E ,  PIROLLI PETER L ,  CHI ED H ,  LI JUN

IPC: G06F17/30 ,  G06F12/00

CPC classification number: G06F17/30994 ,  G06F17/3025 ,  G06F17/3071

Abstract: PROBLEM TO BE SOLVED: To perform document recommendation from documents in collection based on multi-modal user clusters. SOLUTION: An initial set of users are identified, the documents in the collection accessed by the users are identified, the contents of the accessed documents are estimated from the accessed documents for the users, and the users are clustered into a plurality of user clusters by expressing the users using the contents of the documents. A new user is identified, information on the documents accessed by the new user is collected, the contents of the documents accessed by the new user is estimated from the documents accessed by the new user, and the new user is assigned to the user clusters based on similarity between the contents of the documents accessed by the new user and the contents of the documents accessed by other users included in the user clusters. COPYRIGHT: (C)2011,JPO&INPIT

Abstract translation: 要解决的问题：从基于多模式用户群集的收集文档中执行文档推荐。

解决方案：识别初始用户组，识别由用户访问的集合中的文档，从被访问的文档中为用户估计访问文档的内容，并将用户聚类成多个 用户集群通过表达用户使用文档的内容。 识别新的用户，收集新用户访问的文档的信息，由新用户访问的文档估计新用户访问的文档的内容，并将新用户分配给基于用户的用户群 在新用户访问的文档的内容与包含在用户集群中的其他用户访问的文档的内容之间的相似性。 版权所有（C）2011，JPO＆INPIT

公开(公告)号：JP2010218579A

公开(公告)日：2010-09-30

申请号：JP2010140224

申请日：2010-06-21

Applicant: Xerox Corp ,  ゼロックス コーポレイションXerox Corporation

Inventor： CHEN FRANCINE R ,  SCHUETZE HINRICH ,  GARGI ULLAS ,  PITKOW JAMES E ,  PIROLLI PETER L ,  CHI ED H ,  LI JUN

IPC: G06F17/30 ,  G06F12/00

CPC classification number: G06F17/30994 ,  G06F17/3025 ,  G06F17/3071

Abstract: PROBLEM TO BE SOLVED: To visualize a cluster of users represented by a document selected from a collection of documents. SOLUTION: A plurality of users selected in a user group are identified, and the plurality of users share interest decided through multi-mode collection use analysis. For each user of the plurality of users, a corresponding access probability representing the frequency of user's access to documents for each document in a document collection is decided. For each document in the document collection, a set access probability in the plurality of selected users corresponding to the probability that a user in a plurality of selected users accesses the document is computed. A disk tree having a plurality of nodes each representing a document in the document collection is displayed. Each node in the disk tree having a set access probability larger than a desired threshold value is highlighted. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：可视化由从文档集合中选择的文档表示的用户群集。

解决方案：识别在用户组中选择的多个用户，并且多个用户分享通过多模式收集使用分析确定的兴趣。 对于多个用户的每个用户，确定表示用户对文档集合中的每个文档的文档的访问频率的相应访问概率。 对于文档收集中的每个文档，计算与多个选定用户中的用户访问该文档相对应的多个所选用户中的一组访问概率。 显示具有多个节点的磁盘树，每个节点表示文档集合中的文档。 具有大于期望阈值的设定访问概率的盘树中的每个节点被突出显示。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2000339350A

公开(公告)日：2000-12-08

申请号：JP2000016705

申请日：2000-01-26

Applicant: XEROX CORP

Inventor： CHEN FRANCINE R ,  SCHUETZE HINRICH ,  GARGI ULLAS ,  PITKOW JAMES E ,  PIROLLI PETER L ,  CHI ED H ,  LI JUN

IPC: G06F12/00 ,  G06F17/30

Abstract: PROBLEM TO BE SOLVED: To recommend a preferable document to a user by clustering a second object by means of a result obtained by means of separating and searching a feature corresponding to the first plural objects in a collection. SOLUTION: The training set of users is identified from the collection of data (2510) and all types of information being usable concerning the users are collected. Then the users are clustered with multi-mode information in a selection related to multi-mode clustering (2512). In this case, unless the new user exists (2514), the processing is ended. Or when the new user is identified (2518), browsing information is collected from the new user (2520) and the user is assigned to the nearest existing cluster (2522). Then the most popular page in the nearest cluster is identified (2524) and it is recommended to the new user (2526).

公开(公告)号：JP2010267277A

公开(公告)日：2010-11-25

申请号：JP2010140223

申请日：2010-06-21

Applicant: Xerox Corp ,  ゼロックス コーポレイションXerox Corporation

Inventor： CHEN FRANCINE R ,  SCHUETZE HINRICH ,  GARGI ULLAS ,  PITKOW JAMES E ,  PIROLLI PETER L ,  CHI ED H ,  LI JUN

IPC: G06F17/30 ,  G06F12/00

CPC classification number: G06F17/30994 ,  G06F17/3025 ,  G06F17/3071

Abstract: PROBLEM TO BE SOLVED: To select a set of initial cluster centers in wavefront clustering of objects in a collection.
SOLUTION: This method for selecting the set of initial cluster centers selects the set of initial cluster centers in the wavefront clustering of the objects in the collection, wherein each object is represented by a set of vectors with multi-modal features. A first number of a first object is selected from the objects in the collection, the vector centroid of the first object is calculated using the set of vectors with multi-modal features associated with each object, and a second number of a second object is selected from the objects in the collection. The second number of the initial cluster centers between the centroid and the second object is identified. The wavefront clustering is performed to the objects in the collection using the second number of the initial cluster centers.
COPYRIGHT: (C)2011,JPO&INPIT

Abstract translation: 要解决的问题：在集合中的对象的波前聚类中选择一组初始聚类中心。

解决方案：用于选择初始聚类中心集的方法在集合中的对象的波前聚类中选择一组初始聚类中心，其中每个对象由具有多模态特征的一组向量表示。 从集合中的对象中选择第一对象的第一数量，使用具有与每个对象相关联的多模式特征的向量集来计算第一对象的向量中心，并且选择第二对象的第二数量 从集合中的对象。 识别质心和第二个物体之间的第二个初始聚类中心数。 使用第二数量的初始聚类中心对集合中的对象执行波前聚类。 版权所有（C）2011，JPO＆INPIT

公开(公告)号：JP2010250849A

公开(公告)日：2010-11-04

申请号：JP2010140222

申请日：2010-06-21

Applicant: Xerox Corp ,  ゼロックス コーポレイションXerox Corporation

Inventor： CHEN FRANCINE R ,  SCHUETZE HINRICH ,  GARGI ULLAS ,  PITKOW JAMES E ,  PIROLLI PETER L ,  CHI ED H ,  LI JUN

IPC: G06F17/30 ,  G06F12/00

CPC classification number: G06F17/30994 ,  G06F17/3025 ,  G06F17/3071

Abstract: PROBLEM TO BE SOLVED: To calculate a degree of similarity between two objects in the collection of objects. SOLUTION: Two objects in collection which are related with a first feature vector and a second feature vector have two or more dimensions. The first feature vector expresses a first feature of the object, and the second feature vector expresses a second feature of the object. The first feature is a text feature, and the second feature is an image feature. The first feature vector of the first object and the first feature vector of the second object are identified, and the first distance metric between these first feature vectors is calculated. The second feature vector of the first object and the second feature vector of the second object are identified, and the second distance metric between these second feature vectors is calculated. The total of the first distance metric and the second distance metric is calculated. COPYRIGHT: (C)2011,JPO&INPIT

Abstract translation: 要解决的问题：计算物体集合中两个物体之间的相似度。

解决方案：与第一特征向量和第二特征向量相关的集合中的两个对象具有两个或更多个维度。 第一特征向量表示对象的第一特征，第二特征向量表示对象的第二特征。 第一个功能是文本功能，第二个功能是图像功能。 识别第一对象的第一特征向量和第二对象的第一特征向量，并且计算这些第一特征向量之间的第一距离度量。 识别第二物体的第一物体的第二特征向量和第二物体的第二特征向量，并且计算这些第二特征向量之间的第二距离度量。 计算第一距离度量和第二距离度量的总和。 版权所有（C）2011，JPO＆INPIT

公开(公告)号：JP2010205306A

公开(公告)日：2010-09-16

申请号：JP2010140221

申请日：2010-06-21

Applicant: Xerox Corp ,  ゼロックス コーポレイションXerox Corporation

Inventor： CHEN FRANCINE R ,  SCHUETZE HINRICH ,  GARGI ULLAS ,  PITKOW JAMES E ,  PIROLLI PETER L ,  CHI ED H ,  LI JUN

IPC: G06F17/30 ,  G06F12/00

CPC classification number: G06F17/30994 ,  G06F17/3025 ,  G06F17/3071

Abstract: PROBLEM TO BE SOLVED: To represent digital documents in a vector space by using numerical numbers. SOLUTION: A first digital document to be processed is identified from a plurality of digital documents, and a first characteristic provided with a text that encloses images included in the digital document and is not anchor text, and corresponding to the first digital document is extracted from the plurality of digital documents. The first characteristic is converted into a first vector, and the first vector is associated with the first digital document. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：使用数字表示向量空间中的数字文档。 解决方案：从多个数字文档中识别要处理的第一数字文档，以及提供包含文本的第一特征，该文本包含数字文档中包含的图像，而不是锚文本，并且对应于第一数字文档 从多个数字文档中提取出来。 第一特征被转换成第一矢量，第一矢量与第一数字文档相关联。 版权所有（C）2010，JPO＆INPIT