Document clustering method and system

• US 6,728,932 B1
• Filed: 03/22/2000
• Issued: 04/27/2004
• Est. Priority Date: 03/22/2000
• Status: Expired due to Term

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Sayal, Mehmet Oguz, Chundi, Parvathi, Hsu, Meichun, Dayal, Umeshwar
• Primary Examiner(s): Feild, Joseph H.
• Assistant Examiner(s): Blackwell, James H
• Application Number:
  US09/532,539
• Time in Patent Office: 1,497 Days
• Field of Search: 707/2-4, 707/104.1, 715/500, 715/513, 715/514
• US Class Current: 715/230
• CPC Class Codes: G06F 16/355 Class or cluster creation or modification
• CPC Class Codes: Y10S 707/99932 Access augmentation or optimizing
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing

Abstract

Document clustering method and system utilizing both the log-based clustering method and the content-based clustering method are disclosed. The method includes the steps of generating log-based document clusters and combining vectors from the log-based document clusters with individual document clusters for content-based clustering analysis. The log-based document clusters are generated by accessing the retrieval session log, clustering the retrieval sessions, and combining the documents opened during each of the sessions of session clusters.

Claims

1. A method for clustering documents comprising:

generating a hybrid matrix of vectors comprising a first vector representing a first document and a second vector representing a log-based document cluster; and

clustering the documents using the hybrid matrix, wherein the step of generating the hybrid matrix comprises;

accessing retrieval session logs;

clustering retrieval sessions into session clusters;

generating, for each session cluster, a log-based document cluster by combining all documents opened during any retrieval session of the session cluster;

generating a log-based document cluster vector for each of the log-based document clusters;

replacing each document in the log-based document cluster with the log-based document cluster vector;

generating an individual document vector for each document not opened during any retrieval session; and

combining the log-based document cluster vector and the individual document cluster vector, wherein the log-based document cluster is formed by concatenating the documents to be combined.

2. The method recited in claim 1 wherein the second vector is used in place of a second document within the hybrid matrix wherein the second document forms a portion of the log-based document cluster.

3. The method recited in claim 1 wherein the step of clustering the documents using the hybrid matrix is performed using a content-based clustering technique.

4. The method recited in claim 1 wherein the step of clustering retrieval sessions into session clusters comprises the steps of:

generating a Boolean session vector for each retrieval session;

forming a matrix of the Boolean session vectors; and

applying a clustering algorithm to the matrix of the Boolean session vectors.