System and method for event summarization using observer social media messages

1. A method of processing messages pertaining to an event comprising:

receiving, with a network device in a summarization system, a plurality of input messages generated by a plurality of electronic communication devices from a messaging service;

filtering, with the processor in the summarization system, the plurality of input messages based on a keyword filter to identify a plurality of messages pertaining to the event based on messages generated by a plurality of electronic communication devices associated with a plurality of observers of the event;

generating, with the processor in the summarization system, a first message stream that includes a first plurality of messages selected from the plurality of messages pertaining to the event corresponding to a first participant in the event;

generating, with the processor in the summarization system, a second message stream that includes a second plurality of messages selected from the plurality of messages pertaining to the event corresponding to a second participant in the event, the second participant being different than the first participant, at least one message in the first plurality of messages not being in the second plurality of messages and at least one message in the second plurality of messages not being in the first plurality of messages;

identifying, with the processor in the summarization system, a first sub-event in the first message stream with reference to a time distribution and a content distribution of the first plurality of messages in the first message stream, the identification of the first sub-event further comprising:

identifying, with the processor in the summarization system, a first sub-event time corresponding to an increase in message frequency in the first message stream;

generating, with the processor in the summarization system, a first estimated mixture model of the first sub-event with a Gaussian distribution corresponding to the first sub-event time and a multinomial distribution corresponding to content of messages at the first sub-event time; and

generating, with the processor in the summarization system, the first estimated mixture model for the first sub-event with a maximum likelihood estimation (MLE) process applied to the first estimated mixture model with reference to the plurality of messages in the first message stream;

identifying, with the processor in the summarization system, a second sub-event in the second message stream with reference to a time distribution and a content distribution of the second plurality of messages in the second message stream, the identification of the second sub-event further comprising:

identifying, with the processor in the summarization system, a second sub-event time corresponding to an increase in message frequency in the second message stream;

generating, with the processor in the summarization system, a second estimated mixture model of the second sub-event with a Gaussian distribution corresponding to the second sub-event time and a multinomial distribution corresponding to content of messages at the second sub-event time; and

generating, with the processor in the summarization system, the second estimated mixture model for the second sub-event with the MLE process applied to the second estimated mixture model with reference to the plurality of messages in the second message stream;

identifying, with the processor in the summarization system, a global sub-event corresponding to the first sub-event and the second sub-event in response to both the first sub-event time and the second sub-event time occurring within a predetermined time period;

generating, with the processor in the summarization system, a sub-event summary with reference to the global sub-event, the sub-event summary being based on the first plurality of messages in the first message stream that are associated with the first sub-event and the second plurality of messages in the second message stream that are associated with the second sub-event; and

transmitting, with the network device in the summarization system, the sub-event summary to a plurality of electronic communication devices associated with a plurality of users who are not observers of the event.

2. The method of claim 1 , the identification of the global sub-event further comprising:

identifying, with the processor in the summarization system, a level of similarity between the first plurality of messages that are associated with the first sub-event and the second plurality of messages that are associated with the second sub-event; and

identifying the global sub-event corresponding to the first sub-event and the second sub-event in response to the identified level of similarity exceeding a predetermined threshold.

3. The method of claim 1 , the generation of the first message stream further comprising:

identifying, with the processor in the summarization system, a reference to the first participant in at least one message in the plurality of messages pertaining to the event; and

generating, with the processor in the summarization system, a cluster of messages including the at least one message and at least one additional message in the plurality of messages pertaining to the event that have a level of similarity to the at least one message that exceeds a predetermined threshold, the cluster of messages being included in the first plurality of messages.

4. The method of claim 3 , the generation of the cluster of messages further comprising an agglomerative clustering process.

5. The method of claim 1 , the identification of the first sub-event further comprising:

identifying, with the processor in the summarization system, a portion of the first plurality of messages in the first message stream, the portion including at least two messages in the first plurality of messages, with reference to a sub-event mixture model including a Gaussian time distribution and a multinomial content distribution for the portion of the messages that correspond to the first sub-event.

6. The method of claim 1 further comprising:

identifying, with the processor in the summarization system, a third sub-event time corresponding to another increase in message frequency in the first message stream;

generating, with the processor in the summarization system, a third estimated mixture model of the third sub-event with another Gaussian distribution corresponding to the third sub-event time and another multinomial distribution corresponding to content of messages at the third sub-event time;

identifying, with the processor in the summarization system, that a sub-event does not occur at the third sub-event time in response to the MLE process applied to the third estimated mixture model generating a Gaussian distribution with a parameter corresponding to standard of error exceeding a predetermined threshold.

7. The method of claim 1 further comprising:

identifying, with the processor in the summarization system, a third sub-event time corresponding to another increase in message frequency in the first message stream;

generating, with the processor in the summarization system, a third estimated mixture model of the third sub-event with another Gaussian distribution corresponding to the third sub-event time and another multinomial distribution corresponding to content of messages at the third sub-event time;

identifying, with the processor in the summarization system, that the third estimated mixture model corresponds to the first sub-event time in response to the MLE process applied to the third estimated mixture model generating a Gaussian distribution with a parameter corresponding to a mean time being within a predetermined threshold of a mean time parameter of the Gaussian distribution in the first estimated mixture model; and

merging the first estimated mixture model and the third estimated mixture model.

8. A message summarization system comprising:

a memory configured to store programmed instructions and a plurality of messages received from a plurality of electronic communication devices associated with a plurality of observers of an event;

a network device configured to receive messages from a messaging service and to transmit summarization data to a plurality of electronic communication devices associated with a plurality of users who are not observers of the event; and

a processor operatively connected to the memory and the network device, the processor being configured to:

receive a plurality of input messages from the messaging service, the plurality of input messages being generated by a plurality of electronic communication devices;

filter the plurality of input messages based on a keyword filter to identify a plurality of messages pertaining to the event based on messages generated by a plurality of electronic communication devices associated with a plurality of observers of the event;

generate a first message stream that includes a first plurality of messages selected from the plurality of messages pertaining to the event corresponding to a first participant in the event;

identify a first sub-event in the first message stream with reference to a time distribution of messages and content distribution of messages in the first message stream, the processor being further configured to:

identify a first sub-event time corresponding to an increase in message frequency in the first message stream;

generate a first estimated mixture model of the first sub-event with a Gaussian distribution corresponding to the first sub-event time and a multinomial distribution corresponding to content of messages at the first sub-event time; and

generate the sub-event mixture model with a maximum likelihood estimation (MLE) process applied to the first estimated mixture model with reference to the plurality of messages in the first message stream;

generate a second message stream that includes a second plurality of messages selected from the plurality of messages pertaining to the event corresponding to a second participant in the event, the second participant being different than the first participant, at least one message in the first plurality of messages not being in the second plurality of messages and at least one message in the second plurality of messages not being in the first plurality of messages;

identify a second sub-event in the second message stream with reference to a time distribution of messages and content distribution of messages in the second message stream, the processor being further configured to:

identify a second sub-event time corresponding to an increase in message frequency in the second message stream;

generate a second estimated mixture model of the second sub-event with a Gaussian distribution corresponding to the second sub-event time and a multinomial distribution corresponding to content of messages at the second sub-event time; and

generate the second estimated mixture model for the second sub-event with the MLE process applied to the second estimated mixture model with reference to the plurality of messages in the second message stream;

identify a global sub-event corresponding to the first sub-event and the second sub-event in response to both the first sub-event time and the second sub-event time occurring within a predetermined time period;

generate a sub-event summary with reference to the global sub-event, the sub-event summary being based on the first plurality of messages in the first message stream that are associated with the first sub-event and the second plurality of messages in the second message stream that are associated with the second sub-event; and

transmit the sub-event summary to the plurality of electronic communication devices associated with the plurality of users who are not observers of the event.

9. The system of claim 8 , the processor being further configured to:

identify a level of similarity between the first plurality of messages that are associated with the first sub-event and the second plurality of messages that are associated with the second sub-event; and

identify the global sub-event corresponding to the first sub-event and the second sub-event in response to the identified level of similarity exceeding a predetermined threshold.

10. The system claim 8 , the processor being further configured to:

identify a reference to the first participant in at least one message in the plurality of messages pertaining to the event; and

generate a cluster of messages including the at least one message and at least one additional message in the plurality of messages pertaining to the event that have a level of similarity to the at least one message that exceeds a predetermined threshold, the cluster of messages being included in the first plurality of messages.

11. The system of claim 10 , the processor being further configured to perform an agglomerative clustering process to generate the cluster of messages.

12. The system of claim 8 , the processor being further configured to:

identify a portion of the first plurality of messages in the first message stream, the portion including at least two messages in the first plurality of messages, with reference to a sub-event mixture model including a Gaussian time distribution and a multinomial content distribution for the portion of the messages that correspond to the first sub-event.

13. The system of claim 8 , the processor being further configured to:

identify a third sub-event time corresponding to another increase in message frequency in the first message stream;

generate a third estimated mixture model of the third sub-event with another Gaussian distribution corresponding to the third sub-event time and another multinomial distribution corresponding to content of messages at the third sub-event time;

identify that a sub-event does not occur at the third sub-event time in response to the MLE process applied to the third estimated mixture model generating a Gaussian distribution with a parameter corresponding to standard of error exceeding a predetermined threshold.

14. The system of claim 8 , the processor being further configured to:

identify a third sub-event time corresponding to another increase in message frequency in the first message stream;

generate a third estimated mixture model of the third sub-event with another Gaussian distribution corresponding to the third sub-event time and another multinomial distribution corresponding to content of messages at the third sub-event time;

identify that the third estimated mixture model corresponds to the first sub-event time in response to the MLE process applied to the third estimated mixture model generating a Gaussian distribution with a parameter corresponding to a mean time being within a predetermined threshold of a mean time parameter of the Gaussian distribution in the first estimated mixture model; and

merge the first estimated mixture model and the third estimated mixture model.