System and method for interactive multi-resolution topic detection and tracking

• US 10,061,867 B2
• Filed: 12/30/2014
• Issued: 08/28/2018
• Est. Priority Date: 12/30/2014
• Status: Active Grant

• Current Assignee: Genesys Telecommunications Laboratories Incorporated (Genesys Cloud Services Incorporated)
• Original Assignee: Genesys Telecommunications Laboratories Incorporated (Genesys Cloud Services Incorporated)
• Inventors: Lev, Yoni, Faizakof, Avraham, Lev-Tov, Amir, Tapuhi, Tamir, Konig, Yochai
• Primary Examiner(s): Le, Debbie
• Application Number:
  US14/586,730
• Publication Number:
  US 20160188672A1
• Time in Patent Office: 1,337 Days
• CPC Class Codes: G06F 16/358 Browsing; Visualisation therefor
• CPC Class Codes: G06F 16/90332 Natural language query formulation or dialogue systems
• CPC Class Codes: G06Q 30/0281 Customer communication at a business location, e.g. providing product or service information, consulting
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

A method for tracking known topics in a plurality of interactions includes: extracting, by a processor, a plurality of fragments from the plurality of interactions; initializing, by the processor, a collection of tracked topics to an empty collection; computing, by the processor, a similarity between each fragment of the fragments and each of the known topics; and adding, by the processor, a known topic of the known topics to the tracked topics in response to the similarity between a fragment and the known topic exceeding a threshold value.

Claims

1. A method for tracking known topics in a plurality of interactions, the method comprising:

extracting, by a processor, a plurality of fragments from the plurality of interactions, the plurality of interactions occurring over a particular time period, each fragment of the plurality of fragments comprising one or more words;

initializing, by the processor, a collection of tracked topics to an empty collection;

computing, by the processor, a similarity between each fragment of the fragments and each of the known topics, each of the known topics comprising a template fragment comprising one or more template words, the similarity between a fragment of the fragments and a topic of the known topics being computed based on the one or more words of the fragment and the one or more template words of the template fragment of the topic;

adding, by the processor, a known topic of the known topics to the tracked topics in response to the similarity between a fragment and the known topic exceeding a threshold value; and

returning, by the processor, a collection of the tracked topics detected in the plurality of interactions, the collection comprising indications of frequencies at which the tracked topics occur in the plurality of interactions during the particular time period.

2. The method of claim 1, wherein the known topics correspond to a first time period, and

wherein the interactions correspond to a second time period comprising the particular time period, the second time period being different from the first time period.

3. The method of claim 2, further comprising:

receiving a time point from a user interface,

wherein the first time period is before the time point, and

wherein the second time period is after the time point.

4. The method of claim 1, wherein the plurality of interactions correspond to a first time period comprising the particular time period, and

wherein the method further comprises:

extracting, by a processor, a second plurality of fragments from a second plurality of interactions, the second plurality of interactions corresponding to a second time period, the second time period being different from the first time period;

initializing, by the processor, a second collection of tracked topics to the empty collection;

computing, by the processor, a similarity between each fragment of the second fragments and each of the known topics; and

adding, by the processor, a known topic of the known topics to the second tracked topics in response to the similarity between a fragment and the known topic exceeding the threshold value.

5. The method of claim 4, wherein the first time period has a length different from a length of the second time period.

6. The method of claim 4, wherein the first time period has a start time different from a start time of the second time period.

7. The method of claim 1, further comprising:

generating, by the processor, a visualization of the tracked topics.

8. The method of claim 7, further comprising receiving, by the processor, a request to edit a topic of the tracked topics, the request comprising one or more of:

a request to change a color of the topic in the visualization;

a request to change a label of the topic; and

a request to remove fragments from the topic.

9. The method of claim 1, further comprising:

receiving, by the processor, a user request to combine two topics of the tracked topics; and

merging, by the processor, the two topics of the tracked topics.

10. A method for detecting new topics in a plurality of interactions, given a plurality of known topics, the method comprising:

extracting, by a processor, a plurality of fragments from the plurality of interactions, the plurality of interactions occurring over a particular time period, each fragment of the plurality of fragments comprising one or more words;

initializing, by the processor, a collection of detected new fragments to an empty collection;

computing, by the processor, a similarity between each fragment and each of the known topics, each of the known topics comprising a template fragment comprising one or more template words, the similarity between a fragment of the fragments and a topic of the known topics being computed based on the one or more words of the fragment and the one or more template words of the template fragment of the topic;

identifying, by the processor, for each fragment, a corresponding topic having a highest similarity;

adding, by the processor, a fragment of the fragments to the collection of detected new fragments in response to the similarity between the fragment and the corresponding topic having the highest similarity is less than a threshold value;

extracting, by the processor, one or more new topics from the collection of detected new fragments; and

returning, by the processor, a collection of the one or more new topics detected in the plurality of interactions, the collection comprising indications of frequencies at which the one or more new topics occur in the plurality of interactions during the particular time period.

11. The method of claim 10, further comprising:

receiving, by the processor, a request to stop tracking a blacklisted topic;

adding, by the processor, the blacklisted topic to a collection of blacklisted topics; and

computing, by the processor, a similarity between each fragment and each of the blacklisted topics,

wherein the computing, by the processor, the similarity between each fragment and each of the known topics is only performed in response to the similarity of the fragment to any blacklisted topic being less than a blacklist threshold value.

12. The method of claim 10, further comprising:

receiving, by the processor, a request to start tracking a user topic;

adding, by the processor, the user topic to a collection of user topics;

determining, by the processor, whether one or more of the user topics are user suggested topics;

generating, by the processor, a topic for each of the user suggested topics; and

adding the collection of user topics to the plurality of known topics.

13. The method of claim 10, further comprising:

receiving, by the processor, a plurality of blacklisted topics; and

removing, by the processor, all of the blacklisted topics from the one or more new topics.

14. A method for detecting one or more events in a portion of an interaction, given a plurality of known topics, the method comprising:

extracting, by a processor, a plurality of fragments from the portion of the interaction, each fragment of the plurality of fragments comprising one or more words;

initializing, by the processor, a collection of detected new fragments to an empty collection;

computing, by the processor, a similarity between each fragment and each of the known topics, each of the known topics comprising a template fragment comprising one or more template words, the similarity between a fragment of the fragments and a topic of the known topics being computed based on the one or more words of the fragment and the one or more template words of the template fragment of the topic;

identifying, by the processor, for each fragment, a corresponding topic having a highest similarity;

adding, by the processor, a fragment of the fragments to the collection of detected new fragments in response to the similarity between the fragment and the corresponding topic having the highest similarity being greater than a threshold value;

extracting, by the processor, one or more noun phrases from the collection of detected new fragments;

filtering, by the processor, one or more events from the one or more noun phrases; and

returning, by the processor, a collection of the one or more events detected in the portion of the interaction.

15. The method of claim 14, further comprising:

receiving, by the processor, a plurality of blacklisted topics; and

removing, by the processor, all of the blacklisted topics from the one or more events.

16. A method for automatically generating a new topic based on a search query, the method comprising:

receiving, by a processor, the search query;

determining, by the processor, a plurality of fragments corresponding to the search query, each fragment of the plurality of fragments comprising one or more words;

identifying, by the processor, a template fragment of the plurality of fragments, the template fragment comprising one or more template words, each of the fragments having a similarity to the template fragment exceeding a threshold, each similarity between a fragment of the fragments and the template fragment being computed based on the one or more words of the fragment and the one or more template words of the template fragment; and

generating, by the processor, the new topic in accordance with the template fragment and the fragments.

17. A system for tracking known topics in a plurality of interactions, the system comprising:

a processor; and

memory, wherein the memory has stored thereon instructions that, when executed by the processor, cause the processor to;

extract a plurality of fragments from the plurality of interactions, the plurality of interactions occurring over a particular time period, each fragment of the plurality of fragments comprising one or more words;

initialize a collection of tracked topics to an empty collection;

compute a similarity between each fragment of the fragments and each of the known topics, each of the known topics comprising a template fragment comprising one or more template words, the similarity between a fragment of the fragments and a topic of the known topics being computed based on the one or more words of the fragment and the one or more template words of the template fragment of the topic;

add a known topic of the known topics to the tracked topics in response to the similarity between a fragment and the known topic exceeding a threshold value; and

return a collection of the tracked topics detected in the plurality of interactions, the collection comprising indications of frequencies at which the tracked topics occur in the plurality of interactions during the particular time period.

18. The system of claim 17, wherein the known topics correspond to a first time period, and

wherein the interactions correspond to a second time period, the second time period being different from the first time period.

19. The system of claim 18, wherein the memory further stores instructions that, when executed by the processor, cause the processor to:

receive a time point from a user interface,

wherein the first time period is before the time point, and

wherein the second time period is after the time point.

20. The system of claim 17, wherein the plurality of interactions correspond to a first time period, and

wherein the memory further stores instructions that, when executed by the processor, cause the processor to:

extract a second plurality of fragments from a second plurality of interactions, the second plurality of interactions corresponding to a second time period, the second time period being different from the first time period;

initialize a second collection of tracked topics to the empty collection;

compute a similarity between each fragment of the second fragments and each of the known topics; and

add a known topic of the known topics to the second tracked topics in response to the similarity between a fragment and the known topic exceeding the threshold value.

21. The system of claim 20, wherein the first time period has a length different from a length of the second time period.

22. The system of claim 20, wherein the first time period has a start time different from a start time of the second time period.

23. The system of claim 17, wherein the memory further stores instructions that, when executed by the processor, cause the processor to:

generate, by the processor, a visualization of the tracked topics.

24. The system of claim 23, wherein the memory further stores instructions that, when executed by the processor, cause the processor to:

receive a request to edit a topic of the tracked topics, the request comprising one or more of;

a request to change a color of the topic in the visualization;

a request to change a label of the topic; and

a request to remove fragments from the topic.

25. The system of claim 17, wherein the memory further stores instructions that, when executed by the processor, cause the processor to:

receive a user request to combine two topics of the tracked topics; and

merge the two topics of the tracked topics.

26. A system for detecting new topics in a plurality of interactions, given a plurality of known topics, the system comprising:

a processor; and

memory, wherein the memory has stored thereon instructions that, when executed by the processor, cause the processor to;

extract a plurality of fragments from the plurality of interactions, the plurality of interactions occurring over a particular time period, each fragment of the plurality of fragments comprising one or more words;

initialize a collection of detected new fragments to an empty collection;

compute a similarity between each fragment and each of the known topics, each of the known topics comprising a template fragment comprising one or more template words, the similarity between a fragment of the fragments and a topic of the known topics being computed based on the one or more words of the fragment and the one or more template words of the template fragment of the topic;

identify, for each fragment, a corresponding topic having a highest similarity;

add a fragment of the fragments to the collection of detected new fragments in response to the similarity between the fragment and the corresponding topic having the highest similarity is less than a threshold value;

extract one or more new topics from the collection of detected new fragments; and

return a collection of the one or more new topics detected in the plurality of interactions, the collection comprising indications of frequencies at which the one or more new topics occur in the plurality of interactions during the particular time period.

27. The system of claim 26, wherein the memory further has stored thereon instructions that, when executed by the processor, cause the processor to:

receive a request to stop tracking a blacklisted topic;

add the blacklisted topic to a collection of blacklisted topics; and

compute a similarity between each fragment and each of the blacklisted topics,

wherein the instructions to compute the similarity between each fragment and each of the known topics are only executed in response to the similarity of the fragment to any blacklisted topic being less than a blacklist threshold value.

28. The system of claim 26, wherein the memory further has stored thereon instructions that, when executed by the processor, cause the processor to:

receive a request to start tracking a user topic;

add the user topic to a collection of user topics;

determine whether one or more of the user topics are user suggested topics;

generate a topic for each of the user suggested topics; and

add the collection of user topics to the plurality of known topics.

29. The system of claim 26, wherein the memory further has stored thereon instructions that, when executed by the processor, cause the processor to:

receive a plurality of blacklisted topics; and

remove all of the blacklisted topics from the one or more new topics.

30. A system for detecting one or more events in a portion of an interaction, given a plurality of known topics, the system comprising:

a processor; and

memory, wherein the memory has stored thereon instructions that, when executed by the processor, cause the processor to;

extract a plurality of fragments from the portion of the interaction, each fragment of the plurality of fragments comprising one or more words;

initialize a collection of detected new fragments to an empty collection;

compute a similarity between each fragment and each of the known topics, each of the known topics comprising a template fragment comprising one or more template words, the similarity between a fragment of the fragments and a topic of the known topics being computed based on the one or more words of the fragment and the one or more template words of the template fragment of the topic;

identify, for each fragment, a corresponding topic having a highest similarity;

add a fragment of the fragments to the collection of detected new fragments in response to the similarity between the fragment and the corresponding topic having the highest similarity being greater than a threshold value;

extract one or more noun phrases from the collection of detected new fragments;

filter one or more events from the one or more noun phrases; and

return a collection of the one or more events detected in the portion of the interaction.

31. The system of claim 30, wherein the memory further has stored thereon instructions that, when executed by the processor, cause the processor to:

receive a plurality of blacklisted topics; and

remove all of the blacklisted topics from the one or more events.

32. A system for automatically generating a new topic based on a search query, the system comprising:

a processor; and

memory, wherein the memory has stored thereon instructions that, when executed by the processor, cause the processor to;

receive the search query;

determine a plurality of fragments corresponding to the search query, each fragment of the plurality of fragments comprising one or more words;

identify a template fragment of the plurality of fragments, the template fragment comprising one or more template words, each of the fragments having a similarity to the template fragment exceeding a threshold, each similarity between a fragment of the fragments and the template fragment being computed based on the one or more words of the fragment and the one or more template words of the template fragment; and

generate the new topic in accordance with the template fragment and the fragments.