Systems and methods for detecting emerging events

1. A computer-implemented method for detecting emerging events comprising:

obtaining current communication data associated with communication between a first plurality of sources during a first time period;

segmenting the current communication data into a set of data segments;

determining, for a data segment, whether the data segment belongs to a familiar topic of a set of familiar topics;

when the data segment is determined to be belonging to the familiar topic:

determining a mention metric of the familiar topic during the first time period;

determining whether the mention metric falls outside of an associated metric range for more than an associated breach threshold; and

generating, upon determining the mention metric falls outside of the associated metric range more than the associated breach threshold, a first notification indicative of a familiar-topic-related emerging event; or

when the data segment is determined to be belonging to none of the set of familiar topics:

determining an unfamiliar segment count for a set of unfamiliar data segments determined as not belonging to the familiar topic, wherein the set of unfamiliar data segments are similar to each other;

determining whether the unfamiliar segment count meets or exceeds a critical mass threshold; and

generating, upon determining the unfamiliar segment count meets or exceeds the critical mass threshold, a second notification indicative of a new-topic-related emerging event.

2. The computer-implemented method of claim 1 , further comprising:

obtaining historic communication data associated with communication between a second plurality of sources during a second time period prior to the first time period; and

identifying the set of familiar topics based at least in part upon the historic communication data.

3. The computer-implemented method of claim 1 , wherein the current communication data are time-stamped and include text communication data or transcribed audio communication data between the plurality of sources.

4. The computer-implemented method of claim 1 , wherein one or more familiar topics of the set of familiar topics are labeled by a set of keywords that are conceptually salient but not textually similar to each other.

5. The computer-implemented method of claim 1 , wherein one or more data segments comprise one or more sequential sentences or a partial sentence.

6. The computer-implemented method of claim 1 , wherein determining whether the data segment belongs to the familiar topic of a set of familiar topics comprises:

transforming the data segment into a current feature vector;

mapping the current feature vector onto a feature space, the feature space mapped with a set of familiar topic feature vectors associated with the set of familiar topics;

for each familiar topic feature vector:

determining a deviation distance between the current feature vector to the familiar topic feature vector; and

determining a similarity score indicative of how similar the current feature vector is to an associated familiar topic; and

determining a highest similarity score that exceeds a similarity score threshold, identifying the data segment as belonging to the familiar topic associated with the highest similarity score; or

upon determining no similarity score exceeds the similarity score threshold, identifying the data segment as belonging to none of the set of familiar topics.

7. The computer-implemented method of claim 6 , wherein the set of unfamiliar data segments are spatially within a critical mass boundary in the feature space.

8. The computer-implemented method of claim 6 , further comprising:

transforming the set of familiar topics into the set of familiar topic feature vectors; and

mapping the set of familiar topic feature vectors onto the feature space.

9. The computer-implemented method of claim 1 , wherein determining the mention metric further comprises:

determining a rate of mention as the mention metric based at least in part upon a first mention count during a first time segment of the first time period; and

repeatedly updating the rate of mention based at least in part upon an immediate subsequent mention count during an immediate subsequent time segment of the first time period until an end of the first time period.

10. The computer-implemented method of claim 9 , wherein:

the rate of mention is a rolling rate of mention where a subsequent time segment overlaps partially with a time segment before it; or

the rate of mention is a non-rolling rate of mention where the subsequent time segment has no overlap with the time segment before it.

11. The computer-implemented method of claim 9 , further comprising:

obtaining historic communication data associated with communication between a second plurality of sources during a second time period prior to the first time period;

identifying the set of familiar topics based at least in part upon the historic communication data using a topic identification model; and

predicting, for one or more familiar topics in the set of familiar topics, a rate range as the associated metric range based at least in part upon the historic communication data using a time-series prediction model;

wherein the rate range includes a lower rate threshold and a higher rate threshold.

12. The computer-implemented method of claim 11 , further comprising:

obtaining a detection sensitivity;

wherein predicting the rate range is further based upon the detection sensitivity such that the rate range is wider when the detection sensitivity is lower and the rate range is narrower when the detection sensitivity is higher.

13. The computer-implemented method of claim 1 , further comprising:

obtaining a detection sensitivity; and

determining, for one or more familiar topics, a count threshold as part of the associated breach threshold based upon the detection sensitivity such that the count threshold is higher when the detection sensitivity is lower and the count threshold is lower when the detection sensitivity is higher.

14. The computer-implemented method of claim 1 , further comprising:

obtaining a detection sensitivity; and

determining, for one or more familiar topics, a level threshold as part of the associated breach threshold based upon the detection sensitivity such that the level threshold is higher when the detection sensitivity is lower and the level threshold is lower when the detection sensitivity is higher.

15. The computer-implemented method of claim 1 , further comprising:

for one or more familiar topics:

recording a rolling breach count associated with a number of instances the mention metric meets or exceeds the associated metric range within a rolling time window; and

updating the rolling breach count when the rolling time window moves forward.

16. The computer-implemented method of claim 1 , further comprising:

transmitting the first notification to a recipient;

wherein the first notification and the second notification comprises a description of the familiar topic, a set of representative communication data available for review, and a mention trend.

17. The computer-implemented method of claim 1 , further comprising:

when the data segment is determined to be belonging to a familiar topic:

generating, upon determining the mention metric falls outside of the associated metric range, a third notification indicative of a potential familiar-topic-related emerging event; or

when the data segment is determined to be belonging to none of the set of familiar topics:

generating a fourth notification indicative of a potential new-topic-related emerging event.

18. The computer-implemented method of claim 1 , wherein:

the mention metric is a mention count and the associated metric range is a mention range;

the mention metric is a rate of mention and the associated metric range is a rate range; or

the mention metric is an acceleration of mention and the associated metric range is an acceleration range.

19. A system for detecting emerging events comprising:

a processor; and

a memory storing computer-executable instructions that, upon execution by the processor, causes the system to perform a set of operations, the set of operations comprising:

obtaining current communication data associated with communication between a first plurality of sources during a first time period;

segmenting the current communication data into a set of data segments;

determining, for a data segment, whether the data segment belongs to a familiar topic of a set of familiar topics;

when the data segment is determined to be belonging to the familiar topic:

determining a mention metric of the familiar topic during the first time period;

determining whether the mention metric falls outside of an associated metric range for more than an associated breach threshold; and

generating, upon determining the mention metric falls outside of the associated metric range more than the associated breach threshold, a first notification indicative of a familiar-topic-related emerging event; or

when the data segment is determined to be belonging to none of the set of familiar topics:

determining an unfamiliar segment count for a set of unfamiliar data segments determined as not belonging to the familiar topic, the set of unfamiliar data segments being similar to each other;

determining whether the unfamiliar segment count meets or exceeds a critical mass threshold; and

generating, upon determining the unfamiliar segment count meets or exceeds the critical mass threshold, a second notification indicative of a new-topic-related emerging event.

20. A computer-implemented method for detecting emerging events comprising:

obtaining current communication data associated with communication between a first plurality of sources during a first time period;

segmenting the current communication data into a set of data segments;

determining, for a data segment, whether the data segment belongs to a familiar topic of a set of familiar topics; and

when the data segment is determined to be belonging to none of the set of familiar topics:

determining an unfamiliar segment count for a set of unfamiliar data segments determined as not belonging to the familiar topic, the set of unfamiliar data segments being similar to each other;

determining whether the unfamiliar segment count meets or exceeds a critical mass threshold; and

generating, upon determining the unfamiliar segment count meets or exceeds the critical mass threshold, a second notification indicative of a new-topic-related emerging event.