Topic navigation in interactive dialog systems

1. A computer-implemented method for dynamically managing an interactive dialogue in a digital environment, comprising:

recursively determining a set of sector points that equally divide a distance between a current topic and a target topic;

determining, after identifying, the current topic and the target topic, a path through a vector space using a hardware processor, from the current topic to the target topic, the path comprising a set or bridging topics based on topic words that are close to the sector points, where each bridging topic in the path is determined based on a semantic similarity score defined by cosine similarity between word-vectors in a vector space, and is within a threshold distance in the vector space from a previous topic and a subsequent topic;

conducting an interactive conversation with a user, including accepting inputs from the user via a user interface and providing responses to the user via the user interface, introducing each topic in the path until the target topic has been reached;

analyzing, using natural language processing (NLP), inputs from the user and responses during the interactive conversation to refine the set of bridging tonics based on contextual relevance; and

recording at least one predetermined piece of information relating to the inputs from the user and the responses during the interactive conversation for the target topic.

2. The method of claim 1 , wherein determining the path from the current tonic to the target topic further comprises determining that the topic words share no common neighbors with the current topic or the target topic.

3. The method of claim 2 , wherein determining the path from the current topic to the target topic further comprises incrementing a value that determines which sector points are used and determining new topic words close to new sector points based on the in value.

4. The method of claim 1 , wherein determining the path from the current topic to the target topic further comprises determining that the topic words both share common neighbors with the current topic or the target topic.

5. The method of claim 4 , wherein determining the path from the current topic to the target topic further comprises identifying topics for the path from the common neighbors shared with the current topic and the target topic.

6. The method of claim 1 , wherein determining the path from the current topic to the target topic comprises determining that the current topic and the target topic share at least one common neighbor.

7. The method of claim 6 , wherein determining the path from the current topic to the target topic comprises identifying a topic for the path from the at least one common neighbor.

8. The method of claim 1 , farther comprising determining a state of the user based on the at least one predetermined piece of information and performing a corrective action based on the determined state of the user.

9. The method of claim 1 , wherein conducting an interactive conversation with the user is performed after determining the path.

10. The method of claim 1 , wherein a neighborhood around a topic in the vector space includes topics related to the topic.

11. A non-transitory computer readable storage medium comprising a computer readable program for dynamically managing an interactive dialogue in a digital environment, wherein the computer readable program when executed on a computer causes the computer to perform steps of:

recursively determining a set of sector points that equally divide a distance between a current topic and a target topic;

determining, after identifying the current topic and the target topic, a path through a semantic vector space, from the current topic to the target topic, the path comprising a set of bridging topics based on topic words that are close to the sector points, where each bridging topic in the path is determined based on a semantic similarity score defined by cosine similarity between word-vectors in a vector space, and is within a threshold distance in the vector space from a previous topic and a subsequent topic;

conducting an interactive conversation with a user, including accepting inputs from the user via a user interface and providing responses to the user via the user interface, introducing each topic in the path until the target topic has been reached;

analyzing, using natural language processing (NLP), inputs from the user and responses during the interactive conversation to refine the set of bridging topics based on contextual relevance; and

recording at least one predetermined piece of information relating to the inputs from the user and the responses during the interactive conversation for the target topic.

12. The non-transitory computer readable storage medium of claim 11 , wherein determining the path from the current topic to the target topic further comprises determining that the topic words share no common neighbors with the current topic or the target topic.

13. The non-transitory computer readable storage medium of claim 12 , wherein determining the path from the current topic to the target topic further comprises incrementing a value that determines which sector points are used and determining new topic words close to new sector points based on the incremented value.

14. An interactive dialog system, comprising:

a hardware processor; and

a memory that stores a computer program, which, when Executed by the Hardware processor, causes the hardware processor to:

recursively determine a set of sector points that equally divide a distance between a current topic and a target topic;

determine, after identifying the current topic and the target topic, a path through a vector space, from the current topic to the target topic, that includes a set of bridging topics based on topic words that are close to the sector points, where each bridging topic in the path is d ermined based on a semantic similarity score defined by cosine similarity between word-vectors in a vector space, and is within a threshold distance in the vector space from a previous topic and a subsequent topic;

conduct an interactive conversation with a user, including accepting inputs front the user via a user interface and providing responses to the user via the user interface, introducing each topic the path until the target topic has been reached;

analyze, using natural language processing (NLP), from the user and responses during the interactive conversation to refine the set of bridging topics base contextual relevance; and

record at least one predetermined piece of information relating to the inputs from the user and the responses during the interactive conversation for the target topic.

15. The system of claim 14 , wherein the computer program further causes the hardware processor to determine whether the topic words share have common neighbors with the current topic or the target topic.

16. The system of claim 15 , wherein the computer program further causes the hardware processor to increment a value that determines which sector points are used and determines new topic words close to new sector points based on the incremented value when the topic words have no common neighbors with the current topic and the target topic.

17. The system of claim 15 , wherein the computer program further causes the hardware processor to identify topics for the path from the common neighbors shared with the current topic and the target topic when the topic words both have common neighbors with the current topic or the target topic.

18. The system of claim 14 , wherein the computer program further causes the hardware processor to identify a topic for the path from a common neighbor when the current topic and the target topic share at least one common neighbor.