METHODS AND APPARATUS FOR RECOGNIZING AND ACTING UPON USER INTENTIONS EXPRESSED IN ON-LINE CONVERSATIONS AND SIMILAR ENVIRONMENTS

• US 20120265819A1
• Filed: 06/21/2012
• Published: 10/18/2012
• Est. Priority Date: 08/05/2010
• Status: Active Grant

• Current Assignee: Genesys Telecommunications Laboratories Incorporated (Genesys Cloud Services Incorporated)
• Original Assignee: Solariat, Inc. (Genesys Cloud Services Incorporated)
• Inventors: Davitz, Jeffrey Eric, McGann, Conor
• Granted Patent: US 8,521,818 B2
• Time in Patent Office: Days
• US Class Current: 709/204
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G06Q 30/00 Commerce

Abstract

Intentions of a user as expressed in an on-line conversation or other user generated content (UGC) are identified and subsequently acted upon, for example by bringing relevant or related content to the attention of the user and/or making others aware of the conversation. The intention signals derived from the UGC may be used to fashion or update a user profile or inform a broader view of the individual associated with the UGC.

Claims

1. A computer-implemented method, comprising:

in response to user generated content (UGC) presented via a network-enabled social media platform, determining, by a matching and decision engine of a processor-based system, an intention type and intention topic of an utterance contained within the UGC;

on the basis of the intention type and intention topic, selecting from among a plurality of possible actions, one or more possible actions to be taken;

scoring, according to an assessment of actionability of the intention topic, the one or more possible actions to be taken; and

taking or not taking one or more of the possible actions according to respective scores therefor.

2. The method of claim 1, wherein the one or more possible actions to be taken comprise one or more of:

presenting candidate creatives via the social media platform at which the UGC was posted, informing one or more others of the UGC; and

creating or updating a user profile to account for the intention type and topic.

3. The method of claim 1, wherein the intention type is identified on the basis of features extracted from the utterance.

4. The method of claim 3, wherein the intention topic is identified on the basis of scored n-grams obtained from the utterance.

5. The method of claim 4, wherein the n-grams obtained from the utterance are scored using rules applicable to the identified intention type.

6. The method of claim 1, further comprising generating tagged n-grams from a tokenized representation of the UGC to produce a feature vector representative of the UGC and classifying the tagged n-grams to determine the intention type.

7. The method of claim 6, wherein the intention topic is identified on the basis of scored ones of the n-grams.

8. A computer-implemented method for determining an actionable intention from user generated content (UGC) and selecting one or more creatives in response thereto, the method comprising:

receiving the UGC at a computer processor-based system, and segmenting the UGC into utterances;

for each subject utterance,

extracting features of the subject utterance to produce a feature vector;

classifying the feature vector to produce an intention type of the subject utterance;

determining an intention topic of the subject utterance;

determining candidate creatives for responding to the UGC according to the intention type and intention topic;

scoring the candidate creatives according to a determined actionability of the UGC; and

delivering one or more of the candidate creatives, the one or more selected according to their respective scores, to a user associated with the UGC.

9. The method of claim 8, wherein the utterances are determined and segmented according to punctuation marks.

10. The method of claim 8, wherein the utterances include some or all of questions, assertions, complaints, and requests for action.

11. The method of claim 8, wherein the utterances are represented using a bag of words model, each position in the feature vector is associated with a token and a value at a given position in the feature vector represents an importance of the associated token within the utterance.

12. The method of claim 11, wherein importance of the associated token is measured according to a number of times that the word token appears in the utterance.

13. The method of claim 11, wherein importance of the associated token is measured according to an inverse of a rate of occurrence of the word token across all instances in the UGC.

14. The method of claim 8, wherein when generating the feature vector of the utterance, words within a title of the UGC are weighted more strongly than words occurring in a body of the UGC.

15. The method of claim 8, wherein the intention topic is determined using a linguistics-based approach to identify key noun phrases in the subject utterance.

16. The method of claim 8, wherein the intention topic is determined by tagging parts of speech elements in the subject utterance, analyzing the subject utterance for n-grams, and using identified n-grams as candidate intention topics.

17. The method of claim 16, wherein the n-grams are determined without considering the intention type.

18. The method of claim 17, further comprising scoring the candidate intention topics according to the intention type and determining on the basis of the scores a most likely intention topic for the subject utterance.

19. A server, comprising a processor and a computer-readable storage device communicatively coupled to the processor, the computer-readable storage device storing computer-readable instructions, which when executed by the processor cause the processor to:

determine, in response to user generated content (UGC) presented via a network-enabled social media platform, an intention type and intention topic of an utterance contained within the UGC;

select, on the basis of the intention type and intention topic and from among a plurality of possible actions, one or more possible actions to be taken;

score, according to an assessment of actionability of the intention topic, the one or more possible actions to be taken; and

take or not take one or more of the possible actions according to respective scores therefor.

20. The server of claim 19, wherein the one or more possible actions to be taken comprise one or more of:

presenting candidate creatives via the social media platform at which the UGC was posted, informing one or more others of the UGC; and

updating a user profile to account for the intention type and topic.

21. The server of claim 19, wherein the intention type is identified on the basis of features extracted from the utterance.

22. The server of claim 19, wherein the intention topic is identified on the basis of scored n-grams obtained from the utterance.

23. The server of claim 22, wherein the n-grams obtained from the utterance are scored using rules applicable to the identified intention type.

24. The server of claim 19, further comprising generating tagged n-grams from a tokenized representation of the UGC to produce a feature vector representative of the UGC and classifying the tagged n-grams to determine the intention type.

25. The server of claim 24, wherein the intention topic is identified on the basis of scored ones of the n-grams.