Message recommendation using word isolation and clustering

• US 9,245,013 B2
• Filed: 10/29/2007
• Issued: 01/26/2016
• Est. Priority Date: 11/27/2000
• Status: Expired due to Fees

• Current Assignee: Quest Software, Inc.
• Original Assignee: Dell Software, Inc. (Dell Technologies Inc.)
• Inventors: Oliver, Jonathan James, Buntine, Wray Lindsay, Roumeliotis, George
• Primary Examiner(s): Ahn, Sangwoo
• Application Number:
  US11/927,450
• Publication Number:
  US 20080189253A1
• Time in Patent Office: 3,011 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/334 Query execution G06F16/335 takes precedence
• CPC Class Codes: G06F 16/355 Class or cluster creation or modification
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: H04L 51/18 Commands or executable codes
• CPC Class Codes: H04L 67/02 based on web technology, e.g. hypertext transfer protocol [HTTP]
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

Network system provides a real-time adaptive recommendation set of documents with a high statistical measure of relevancy to the requestor device. The recommendation set is optimized based on analyzing text of documents of the interest set, categorizing these documents into clusters, extracting keywords representing the themes or concepts of documents in the clusters, and filtering a population of eligible documents accessible to the system utilizing site and or Internet-wide search engines. The system is either automatically or manually invoked and it develops and presents the recommendation set in real-time. The recommendation set may be presented as a greeting, notification, alert, HTML fragment, fax, voicemail, or automatic classification or routing of customer e-mail, personal e-mail, job postings, and offers for sale or exchange.

Claims

1. A method for recommending electronic-mail messages, the method comprising:

assembling electronic-mail messages into an interest set, the electronic-mail messages having been previously viewed and responded to by a user, wherein the electronic-mail messages are assembled into an interest set by an assembly module stored in memory, the assembly module executed by a processor;

isolating words in the interest set, wherein the words are isolated by a pre-processing module stored in memory, the pre-processing module executed by a processor;

grouping electronic-mail messages in the interest set into clusters of electronic-mail messages having a level of similarity amongst the isolated words, wherein the electronic-mail messages are grouped by a clustering module stored in memory, the clustering module executed by a processor;

generating a keyword for each cluster, the keyword identified based on one or more terms identified via natural language understanding and representing the theme of the electronic-mail messages in each cluster, wherein the keyword is not one of the terms identified via natural language understanding, the keywords generated by a keyword module when executed by a processor;

recommending an electronic-mail message received subsequent to the generation of the keywords for further action based upon a ranking of the subsequently received electronic-mail message, the ranking based on a number of times one of the generated keywords of the clusters appears in the subsequently received electronic-mail message, wherein the electronic-mail message is recommended by a recommendation module stored in memory, the recommendation module executed by a processor; and

adding the recommended electronic-mail message to the interest set when the user responds to the recommended electronic-mail message by taking the further action, wherein the electronic-mail message is added by the assembly module.

2. The method of claim 1, wherein the electronic-mail messages are from multiple sources.

3. The method of claim 1, wherein execution of the pre-processing module by the processor results in removal of words that are not useful for distinguishing one electronic-mail message from another.

4. The method of claim 3, wherein the non-useful words removed by execution of the pre-processing module are words common to the language of the electronic-mail message.

5. The method of claim 3, wherein the non-useful words removed by execution of the pre-processing module are words that are insignificant to the recommendation.

6. The method of claim 1, further comprising filtering each cluster based on a client-defined limit on electronic-mail message age and an age associated with each electronic-mail message in each cluster, wherein the electronic-mail messages are filtered by a filtration module, the filtration module executed by a processor.

7. A non-transitory computer-readable storage medium having embodied thereon a program, the program being executable by a processor to perform a method for recommending a message, the method comprising:

assembling electronic-mail messages into an interest set, the electronic-mail messages having been previously viewed and responded to by a user;

isolating words in the interest set;

grouping electronic-mail messages in the interest set into clusters of electronic mail messages having a level of similarity amongst the isolated words;

generating a keyword for each cluster, the keyword identified based on one or more terms identified via natural language understanding and representing the theme of the electronic-mail messages in each cluster, wherein the keyword is not one of the terms identified via natural language understanding;

recommending an electronic-mail message received subsequent to the generation of the keywords for further action based upon a ranking of the subsequently received electronic-mail message, the ranking based on a number of times one of the generated keywords of the clusters appears in the subsequently received electronic-mail message; and

adding the recommended electronic-mail message to the interest set when the user responds to the recommended electronic-mail message by taking the further action.

8. The non-transitory computer-readable storage medium of claim 7, wherein words that are not useful for distinguishing one electronic-mail message from another are removed.

9. The non-transitory computer-readable storage medium of claim 8, wherein the removed non-useful words are words common to the language of the electronic-mail message.

10. The non-transitory computer-readable storage medium of claim 8, wherein the removed non-useful words removed are words that are insignificant to the recommendation.

11. The non-transitory computer-readable storage medium of claim 7, wherein the method further comprises filtering each cluster based on a client-defined limit on electronic-mail message age and an age associated with each electronic-mail message in each cluster.

12. A network device, comprising:

a network interface for receiving electronic-mail messages over a communications network;

a processor; and

memory storing;

an assembly module that is executable by the processor and assembles electronic-mail messages into an interest set, the electronic-mail messages having been previously viewed and responded to by a user,

a pre-processing module that is executable by the processor and isolates words in the interest set,

a clustering module that is executable by the processor and groups electronic-mail messages in the interest set into clusters of electronic-mail messages having a level of similarity amongst the isolated words,

a keyword module that is executable by the processor and generates a keyword for each cluster, the keyword identified based on one or more terms identified via natural language understanding and representing the theme of the electronic-mail messages in each cluster, wherein the keyword is not one of the terms identified via natural language understanding, and

a recommendation module that is executable by the processor and recommends an electronic-mail message received subsequent to the generation of the keywords for further action based upon a ranking of the subsequently received electronic-mail message, the ranking based on a number of times one of the generated keywords of the clusters appears in the subsequently received electronic-mail message, wherein the recommended electronic-mail message is added to the interest set when the user responds to the recommended electronic-mail message by taking the further action.

13. The network device of claim 12, wherein the memory further stores a filtration module that is executable by the processor and filters each cluster based on a client-defined limit on electronic-mail message age and an age associated with each electronic-mail message in each cluster.