Content-based classification

• US 9,317,592 B1
• Filed: 01/29/2013
• Issued: 04/19/2016
• Est. Priority Date: 03/31/2006
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Bowman, Courtney, Cabrera, Nicolas Fernando, Hedayati, Keyon, Hotchkiss, Katherine Marie, Hu, Stephen Tai-Chung, Smith, Jared, Sparrow, Isaac David, Stone, Jeffrey Michael, Trinidad, Juan Bacani, Wiesen, David
• Primary Examiner(s): Spooner, Lamont
• Application Number:
  US13/753,066
• Time in Patent Office: 1,176 Days
• Field of Search: 704/1, 704/9, 704/10, 707706-708
• US Class Current: 1/1
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 40/279 Recognition of textual entities

Abstract

A classifier may include logic to parse incoming content and to compare a key term in the content to stored content related to multiple prior messages, where the stored content is classified with respect to one or more categories. The logic may produce a score for the content based on the comparing, relate the score to one of the one or more categories, and produce a result based on the comparing, producing, or relating.

Claims

1. A system, comprising:

a parser configured to parse a received message to identify one or more terms in the received message; and

a classifier operatively coupled to the parser, wherein the classifier is configured to;

compare the one or more terms parsed from the received message to terms in a plurality of previously received messages, wherein the previously received messages are classified in a data storage as stored classified messages with respect to two or more categories, wherein each category of the two or more categories is within a hierarchy of categories, wherein each level in the hierarchy below a topmost level further refines a category of information that is within an above level, and wherein at least one of the terms in the plurality of previously received messages is associated with at least two categories within the hierarchy of categories;

generate a plurality of scores for the received message based on the comparing, wherein each score of the plurality of scores indicates a probability that the received message is related to a particular category of the two or more categories, wherein a first score of the plurality of scores is associated with a first category of the two or more categories, and wherein a second score of the plurality of scores is associated with a second category of the two or more categories that is different from the first category and that is at a same level within the hierarchy of categories as the first category; and

identify that the received message is associated with the first category based on the first score and the second category based on the second score.

2. The system of claim 1, comprising a responder configured to prepare a particular reply message to a destination, wherein the particular reply message is responsive to the received message and is based on the identified first category and the identified second category and is configured to add a copy of the received message and information in the particular reply message to the classifier, wherein the classifier updates the data storage.

3. The system of claim 1, wherein the classifier is configured to identify the first category using a score of a selected one of the plurality of previously received messages, wherein the selected one of the plurality of previously received messages is related to the first category based on a classification performed on the selected one of the plurality of previously received messages.

4. The system of claim 1, wherein the plurality of scores are generated using Bayesian classification.

5. The system of claim 1, wherein a first reply message has been previously sent to a sender of one of the plurality of previously received messages that is associated with at least one of the first category or the second category.

6. The system of claim 1, wherein the received message includes a particular message, wherein the parser is configured to receive another message, wherein the another message is different from the particular message, wherein the classifier is configured to generate a score for the another message by comparing a term in the another message to the terms in the plurality of previously received messages.

7. The system of claim 6, comprising a selector operatively coupled to the classifier, wherein the selector is configured to:

group the particular message with the another message based on the plurality of scores for the particular message and the score for the another message, and

send a reply message to a sender of the particular message and a sender of the another message based on the grouping.

8. The system of claim 1, comprising a message handler operatively coupled to the parser, wherein the message handler is configured to:

receive a plurality of messages,

identify users associated with one or more of the plurality of messages,

identify a priority associated with a plurality of the identified users, and

output the plurality of messages to the parser in an order that is based on identified priorities of users associated with the messages, wherein the parser is configured to receive and to parse the plurality of messages outputted by the message handler in the order that the messages are outputted by the message handler.

9. The system of claim 1, wherein the order in which the messages are output to the parser is an order that is different from an order in which the message handler receives the messages.

10. The system of claim 1, comprising a message handler operatively coupled to the parser, wherein the message handler is configured to:

receive a plurality of messages in a first order,

identify one or more priority terms associated with one or more of the plurality of messages,

modify, based on the identifying, the first order to obtain a different second order, and

output the plurality of messages to the parser in the second order,

wherein the parser is configured to receive and to parse the plurality of messages outputted by the message handler in the second order.

11. The system of claim 1, comprising a message handler that is configured to:

provide a particular reply message to a sender of the received message,

provide a request for feedback regarding the particular reply message, and

receive the feedback regarding the particular reply message.

12. The system of claim 11, wherein the classifier is configured to use the feedback to identify that the received message is associated with a third category of the two or more categories, wherein the third category is different from the first category and the second category.

13. The system of claim 11, wherein the classifier is configured to use the feedback to identify, based on the received message, another reply message that is different from the particular reply message, and wherein the message handler is configured to provide the particular reply message to the sender of the received message.

14. The system of claim 1, wherein the classifier is configured to:

compare the first score to a first threshold associated with the first category; and

compare the second score to a second threshold associated with the second category, wherein the received message is associated with the first category responsive to the first score exceeding the first threshold and his associated with the second category responsive to the second score exceeding the second threshold.

15. The system of claim 14, wherein the first threshold is different from the second threshold.

16. A method, comprising:

parsing, by one or more devices, a received message to identify one or more terms in the received message;

comparing, by the one or more devices, the one or more terms parsed from the received message to terms in a plurality of previously received messages, wherein the previously received messages are classified in a data storage as stored classified messages with respect to two or more categories, wherein each category of the two or more categories is within a hierarchy of categories, wherein each level in the hierarchy below a topmost level further refines a category of information that is within an above level, and wherein at least one of the terms in the plurality of previously received messages is associated with at least two categories within the hierarchy of categories;

generating, by the one or more devices, a plurality of scores for the received message based on the comparing, wherein each score of the plurality of scores, indicates a probability that the received message is related to a particular category of the two or more categories, wherein a first score of the plurality of scores is associated with a first category of the two or more categories, and wherein a second score of the plurality of scores is associated with a second category of the two or more categories that is different from the first category and that is at a same level within the hierarchy of categories as the first category;

comparing, by the one or more devices, the first score to a first threshold;

identifying, by the one or more devices and based on comparing the first score to the first threshold, that the received message is associated with the first category of the two or more categories;

comparing, by the one or more devices, the second score to a second threshold that is different from the first threshold; and

identifying, by the one or more devices and based on comparing the second score to the second threshold, that the received message is associated with the second category of the two or more categories.

17. The method claim 16, further comprising:

providing, by the one or more devices, a particular reply message to a destination in response to the received message, wherein the particular reply message is based on the identified first category and the identified second category; and

adding a copy of the received message and information in the reply message to the classifier, wherein the classifier updates the data storage.

18. The method of claim 16, wherein the generating of the plurality of scores includes identifying relative Bayesian probabilities of the terms with respect to the first category and the second category.

19. The method of claim 16, comprising determining a normalized aggregate score for the received message.

20. The method of claim 16, comprising providing the particular reply message and at least a portion of the first previously received message to a sender of the received message.

21. The method of claim 16, comprising updating a classification made of the received message based on a classification code.

22. The method of claim 21, comprising storing the received message and one or more of the following:

answer content related to the received message, the classification code, information identifying the first category, and information identifying the second category.

23. The method of claim 16, comprising:

identifying a first received message associated with a first user;

identifying a second received message associated with a second user who is associated with a lower priority than the first user;

identifying one or more priority terms in the second message; and

processing the second message before processing the first message, based on identifying the one or more priority terms in the second message.