Automatic organization of documents through email clustering

• US 20070156732A1
• Filed: 12/29/2005
• Published: 07/05/2007
• Est. Priority Date: 12/29/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Platt, John, Renshaw, Erin, Surendran, Arungunram
• Granted Patent: US 7,765,212 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06Q 10/107 Computer-aided management of electronic mailing [e-mailing]
• CPC Class Codes: H04L 51/00 User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail

Abstract

A system that facilitates organization of emails comprises a clustering component that clusters a plurality of emails and creates topics for emails by assigning key phrases extracted from emails within one or more clusters. An organization component then utilizes the key phrases to organize documents. Furthermore, the organization component can comprise a probability component that determines a probability that a document belongs to a certain topic.

Claims

1. A system that facilitates organization of emails comprising the following computer-executable components:

a clustering component that clusters a plurality of emails and creates topics for emails by assigning key phrases extracted from emails within one or more clusters; and

an organization component that utilizes the key phrases to organize documents within a graphical user interface.

2. The system of claim 1, further comprising a pre-processing component that at least one of filters and downweights one or more words within the emails prior to clustering.

3. The system of claim 1, further comprising a filtering component that filters a cluster defined by the clustering component if a number of emails within the cluster is below a predefined threshold.

4. The system of claim 1, the organization component comprises a probability component that determines a probability that a document belongs to a certain topic.

5. The system of claim 4, the clustering component employs Probabilistic Latent Semantic Analysis to automatically characterize each topic by a distribution of words within clusters.

6. The system of claim 4, the probability component determines a probability that a word belongs to a certain topic.

7. The system of claim 1, further comprising a key phrase extraction component that characterizes the topic by selecting descriptive keywords from a cluster and labeling the cluster with such descriptive keywords.

8. The system of claim 1, the documents include one or more of an email, a word processing document, a spreadsheet, a video file, an audio file, and a digital image.

9. The system of claim 1, further comprising a key phrase extraction component that post-processes the key phrases to refine characterization of at least one topic.

10. The system of claim 1, further comprising:

an analysis component that reviews text in subject lines of the emails and creates a list of noun phrases;

a supplement component that supplements the noun phrases with a set of words that appear in the noun phrases; and

a counting component that counts a number of occurrences of the supplemented noun phrases in a body of an email.

11. The system of claim 1, further comprising a token component that supplies tokens to each email within the plurality of emails corresponding to deliverers of the emails and recipients of the emails, the tokens employed by the clustering component to refine a topic and to separate topics that are lexically similar.

12. The system of claim 1, further comprising an interface component that automatically arranges topics defined by the clustering component and presents such arrangement to a user.

13. The system of claim 12, the interface component comprises a relevance calculator component that calculates a relevance measure of an email with respect to a topic that includes the email, documents within a topic can be arranged within the graphical user interface based at least in part upon the calculated relevance measure.

14. The system of claim 1, further comprising an assignment component that assigns an incoming email to at least one topic based upon a calculated probability of relevance to the topic.

15. A method for automatically associating documents with a topic, comprising:

receiving a plurality of emails;

clustering the plurality of emails into multiple clusters;

performing key phrase extraction upon emails within at least one of the clusters;

characterizing a topic with one or more extracted key phrases; and

automatically organizing documents based at least in part upon the one or more extracted key phrases.

16. The method of claim 15, the act of clustering comprises employing a plurality of clustering actions in sequence.

17. The method of claim 16, the act of clustering further comprises:

representing the plurality of emails using term frequency—

inverse document frequency (tƒ

-idƒ

) vectors of selected words;

employing a cosine distance measure to measure email similarity;

creating cluster initializations by performing agglomerative clustering on a subset of the plurality of emails; and

running K-means using the cluster initializations on each of the plurality of emails.

18. The method of claim 15, further comprising:

receiving a new email; and

automatically associating the email with a pre-existing topic.

19. The method of claim 15, further comprising:

displaying a list of expandable topics upon a graphical user interface, the topics correlate to labeled clusters; and

displaying emails associated with the expandable topics upon selection of at least one of the expandable topics.

20. An automatic email organization system, comprising:

computer-implemented means for clustering multiple emails;

computer-implemented means for extracting key phrases from emails within at least one cluster and labeling the at least one cluster with a subset of the extracted key phrases; and

computer-implemented means for organizing documents within a graphical user interface based at least in part upon the labels.