Computer-based interlocutor understanding using classifying conversation segments

1. A computer-based method to create one or more digital models of interlocutory conversations comprising:

receiving, by a computer processor, conversation text data containing one or more transcribed interlocutory conversations between two or more interlocutor devices;

splitting, by a computer processor, the conversation text data into groups related to at least one conversation ontology using metadata associated with the conversation text data;

identifying, by a computer processor, one or more dominant paths of conversational behavior between the groups according to the metadata; and

creating, by a computer processor, a digital conversation model in a computer-readable memory device containing the conversation behaviors, the metadata, and the identified one or more dominant paths, wherein the computer-readable memory device is not a propagating signal per se;

wherein the splitting comprises, at least in part:

providing, by a computer processor, the conversation text data to an unsupervised conversation processing server; and

receiving, by a computer processor, the groups from the unsupervised conversation processing server.

2. The method of claim 1 wherein the at least one conversation ontology defines the groups comprising at least a greeting group, a topic negotiation group, a topic discussion group, a change/end of topic group, and an end-of-conversation group.

3. The method of claim 2 wherein the groups further comprise a topic repair group.

4. The method of claim 1 wherein each group comprises one or more conversational turns, wherein each conversational turn is associated with an interlocutor device.

5. The method of claim 1 wherein the creating a digital conversation model further comprises:

creating, by a computer processor, a data structure stored in a computer-readable memory device which is not a propagating signal per se;

creating, by a computer processor, in the data structure, at least one top-level topic record, wherein the at least one top-level topic record comprises a plurality of weight values for conversational paths arriving to the topic from at least two previous groups;

wherein the weight values represent historical conversational behaviors leading to the topic and are predictive of future conversational behaviors about the same topic.

6. The method of claim 1 wherein the creating a digital conversation model further comprises:

creating, by a computer processor, a data structure stored in a computer-readable memory device which is not a propagating signal per se;

creating, by a computer processor, in the data structure, at least one top-level topic record, wherein the top level topic record comprises a plurality of weight values for conversational paths departing from a topic to at least two next groups;

wherein the plurality of weight values represent historical conversational behaviors leading away from the topic and are predictive of future conversational behaviors about the same topic.

7. The method as set forth in claim 1 wherein the metadata comprises marks associated with conversational turns which indicate a conversational group to which each conversational turn belongs.

8. The method as set forth in claim 1 wherein the metadata comprises one or counts, frequencies, statistics, or a combination of counts, frequencies and statistics, associated with each group, each topic, and each conversational path between groups in the conversational text data.

9. The method of claim 8 wherein one or more dominant paths of conversational behavior are indicated by counts, frequencies or statistics which exceed a threshold, wherein the one or more dominant paths are paths of conversation which are most expected to lead to or depart from a particular group.

10. The method of claim 1 wherein the received conversational text data comprises transcriptions from one or more sources consisting of an online chat or a text messaging system, a speech recognition system, a chatbot and a voicebot system.

11. The method of claim 1 wherein the splitting comprises, at least in part: providing, by a computer processor, the conversation text data to a human interface device; and receiving, by a computer processor, the groups from the human interface device.

12. The method as set forth in claim 1 wherein the conversation text data comprises a plurality of conversations, wherein at least one common interlocutor is included in all the plurality of conversations.

13. The method as set forth in claim 1 wherein the conversation text data comprises a plurality of conversations, wherein more than two different interlocutors are included within the plurality of conversations.

14. The method as set forth in claim 1 wherein the conversation text data comprises turns from at least one automated interlocutor.

15. A computer program product to create one or more digital models of interlocutory conversations comprising:

a tangible, computer-readable memory device which is not a propagating signal per se; and

program instructions encoded by the tangible, computer-readable memory device which, when executed by a processor, perform:

receiving conversation text data containing one or more transcribed interlocutory conversations between two or more interlocutor devices;

splitting the conversation text data into groups related to at least one conversation ontology using metadata associated with the conversation text data;

identifying one or more dominant paths of conversational behavior between the groups according to the metadata; and

creating a digital conversation model in computer memory containing the conversation behaviors, the metadata, and the identified dominant paths;

wherein the splitting comprises, at least in part:

providing the conversation text data to an unsupervised conversation processing server; and

receiving the groups from the unsupervised conversation processor.

16. The computer program product of claim 15 wherein the conversation ontology defines the groups comprising at least a greeting group, a topic negotiation group, a topic discussion group, a change/end of topic group, and an end-of-conversation group, and wherein each group comprises one or more conversational turns, wherein each turn is associated with an interlocutor device.

17. A system to create one or more digital models of interlocutory conversations

comprising:

a computer processor;

a tangible, computer-readable memory device which is not a propagating signal per se; and

program instructions encoded by the tangible, computer-readable memory device which, when executed by the computer processor, perform:

receiving conversation text data containing one or more transcribed interlocutory conversations between two or more interlocutor devices;

splitting the conversation text data into groups related to at least one conversation ontology using metadata associated with the conversation text data;

identifying one or more dominant paths of conversational behavior between the groups according to the metadata; and

creating a digital conversation model in computer memory containing the conversation behaviors, the metadata, and the identified dominant paths;

wherein the splitting comprises, at least in part:

providing the conversation text data to an unsupervised conversation processing server; and

receiving the groups from the unsupervised conversation processor.

18. The system of claim 17 wherein the conversation ontology defines the groups comprising at least a greeting group, a topic negotiation group, a topic discussion group, a change/end of topic group, and an end-of-conversation group, wherein each group comprises one or more conversational turns, wherein each turn is associated with an interlocutor device.