Topic mining based on interactionally defined activity sequences

1. A computing device comprising:

a processor;

a network interface coupled to the processor to enable communication over a network;

an interaction engine configured to perform acts comprising:

receiving an electronic data package comprising a sequence of utterances between conversation entities, over the network;

classifying each utterance to a corresponding social action;

identifying transition boundaries based on the social actions;

grouping one or more utterances between transition boundaries in the sequence to a segment based on a deep learning model;

determining a similarity of topics between adjacent segments; and

upon determining that the similarity is above a predetermined threshold, grouping the adjacent segments together; and

storing a transcript of the conversation including the grouping of the adjacent segments in the memory.

2. The computing device of claim 1 , wherein the electronic data package comprises raw audio data.

3. The computing device of claim 2 , wherein the interaction engine is further configured to perform acts, comprising: converting a speech of the raw audio data to text by way of natural language processing (NLP).

4. The computing device of claim 2 , wherein the interaction engine is further configured to perform acts, comprising:

identifying a duration of silences between the utterances; and

including each duration in the transcript.

5. The computing device of claim 1 , wherein the classification of each utterance to a corresponding social action comprises using concept expansion to identify the social action.

6. The computing device of claim 1 , wherein each grouping of the one or more utterances into a segment is based on a common focused topic.

7. The computing device of claim 6 , wherein the topic is broadened for each iteration of merging the sequence of adjacent segments.

8. The computing device of claim 1 , wherein each segment comprises a sequence of adjacent utterances between transition boundaries.

9. The computing device of claim 1 , wherein the interaction engine is further configured to perform acts, comprising: removing out utterances that are not interactionally defined to reduce the computational complexity and storage requirements of the computing device.

10. The computing device of claim 1 , wherein the deep learning model is created by the computing device during a training phase, comprising:

receiving historical user interaction logs between conversation entities;

for each interaction log:

using concept expansion for a creation of a specialized dictionary for identifying social actions associated with each utterance in the interaction log;

classifying each utterance into a corresponding social action; and

grouping one or more utterances associated with social actions in a predetermined window in a sequence of the interaction log to a transition boundary.

11. The computing device of claim 10 , wherein the deep learning model is a sequential recurrent neural network (RNN) having a predetermined window.

12. The computing device of claim 1 , wherein the deep learning model is a many to one sequential recurrent neural network (RNN).

13. A non-transitory computer readable storage medium tangibly embodying a computer readable program code having computer readable instructions that, when executed, causes a computer device to carry out a method of determining topics of a conversation, the method comprising:

receiving an electronic data package comprising a sequence of utterances between conversation entities;

classifying each utterance to a corresponding social action;

identifying transition boundaries based on the social actions; grouping one or more utterances between transition boundaries in the sequence to a segment based on a deep learning model;

determining a similarity of topics between adjacent segments; and

upon determining that the similarity is above a predetermined threshold, grouping the adjacent segments together; and

storing a transcript of the conversation including the grouping of the adjacent segments in the memory.

14. The non-transitory computer readable storage medium of claim 13 , wherein the electronic data package comprises raw audio data.

15. The non-transitory computer readable storage medium of claim 14 , further comprising: converting a speech of the raw audio data to text by way of natural language processing (NLP).

16. The non-transitory computer readable storage medium of claim 13 , wherein the classification of each utterance to a corresponding social action comprises using concept expansion to identify the social action.

17. The non-transitory computer readable storage medium of claim 13 , wherein each grouping of the one or more utterances into a segment is based on a common focused topic.

18. The non-transitory computer readable storage medium of claim 17 , wherein the topic the predetermined threshold is replaced with has a more loose threshold for each iteration of merging the sequence of adjacent segments.

19. The non-transitory computer readable storage medium of claim 13 , wherein the deep learning model is created during a training phase, comprising:

receiving historical user interaction logs between conversation entities;

for each interaction log:

using concept expansion for a creation of a specialized dictionary for identifying social actions associated with each utterance in the interaction log;

classifying each utterance of the interaction log into a corresponding social action; and

grouping one or more utterances in a predetermined window in a sequence of the interaction log to a transition boundary.

20. The non-transitory computer readable storage medium of claim 19 , wherein the deep learning model is a many to one sequential recurrent neural network (RNN) having a predetermined window.