System and method for detecting the occurrence of an event and determining a response to the event

Patent metadata

Field	Value
Publication number	US-11934937-B2
Application number	US-201715645757-A
Country	US
Kind code	B2
Filing date	Jul 10, 2017
Priority date	Jul 10, 2017
Publication date	Mar 19, 2024
Grant date	Mar 19, 2024

Abstract

A system for predicting the occurrence of an event includes an event detector and a reporting processor. The event detector is configured to: receive data that defines a plurality of social media items; receive a real-time data feed; and predict the occurrence of an event based on a correlation between information in the plurality of social media items and activity associated with the real-time data feed. The reporting processor is configured to determine an event type associated with the event; identify a sentiment of the predicted event based on historical data in the real-time data feed, and generate a recommendation for preventing the occurrence of the event based on at least one of the event type and the sentiment of the predicted event. The recommendation includes a plurality of actions. The reporting processor is coupled to a knowledge graph database that corresponds to an ontology that defines one or more relationships between event types, and response types. The reporting processor determines an order of the actions of the recommendation based on the knowledge graph ontology.

First claim

We claim:

1. A system comprising:
non-transitory memory storing instructions executable to predict an occurrence of an event; and
a processor configured to execute the instructions to:
receive data that defines a plurality of social media items;
in response to receiving the data, categorize each social media item of the plurality of social media items into each topic of a plurality of topics;
in response to categorizing the each social media item into the each topic, determine whether a plurality of social media items have been categorized to a target topic within a predetermined amount of time according to evaluating time stamps of the social media items, and select the target topic when a predetermined number of social media items have been posted to the target topic;
in response to selecting the target topic, determine a geographic location associated with the predetermined number of social media items included in the target topic;
in response to determining the geographic location, select and receive a real- time data feed associated with the determined geographic location;
in response to receiving the selected real-time data feed, analyze the selected real-time data feed to determine the occurrence of an activity associated with the real-time data feed, and predict the occurrence of the event based on a correlation between information in the plurality of social media items and the activity associated with the real-time data feed;
in response to predicting the occurrence of the event, determine an event type associated with the event by assessing the event type associated with each social media item of the plurality of social media items using distributional compositional semantics and hierarchical language modeling;
in response to determining the event type, identify a sentiment of the event by calculating a sentiment value that is an average value of a plurality of sentiments associated with all of the plurality of social media items;
determine a response type based on the identified sentiment of the event and supporting evidence of the determined event type; and
transmit a message associated with the determined response type to a first responder terminal.

2. The system according to claim 1 , wherein the processor is configured to:
generate vector representations of each word in each of the social media items of the target topic;
generate by a long-short term (LSTM) neural network a vector representation of predicted activities associated with the real-time data feed; and
generate by a feed forward neural network an output indicative of whether the social media items are associated with an event based on the vector representations of each word in each of the social media items of the target topic and the vector representation of predicted activities.

3. The system according to claim 2 , wherein the processor is further configured to query a knowledge graph database that defines one or more relationships between event types, sentiment types, and response types, and wherein the processor is further configured to locate in the knowledge graph the response type associated with the determined sentiment and the event type and generates the recommendation based on the located response type.

4. The system according to claim 1 , wherein the processor is further configured to aggregate the social media items associated with the target topic and to determine a sentiment associated with the aggregate.

5. The system according to claim 1 , wherein the processor is configured to determine the event type based on keywords in the social media items associated with the target topic.

6. The system according to claim 1 , wherein the real-time data feed corresponds to one or more of: a video data feed, an audio data feed, a news data feed, a weather data feed, and historical event data feed.

7. A non-transitory computer readable medium that includes instruction code that facilitates determining the occurrence of an event, the instruction code being executable by a machine for causing the machine to perform acts comprising:
implementing an event detector engine configured to:
receive data that defines a plurality of social media items;
determine a geographic location associated with a predetermined number of social media items included in a target topic;
in response to determining the geographic location, select and receive a real-time data feed associated with the determined geographic location; and

in response to receiving the selected real-time data feed, analyze the selected real-time data feed to determine the occurrence of an activity associated with the real-time data feed, and determine the occurrence of the event based on a correlation between information in the plurality of social media items and the activity associated with the real-time data feed;
implementing a reporting engine configured to:
determine an event type associated with the event by assessing the event type associated with each social media item of the plurality of social media items using distributional compositional semantics and hierarchical language modeling;
in response to determining the event type, identify a sentiment of the event by calculating a sentiment value that is an average value of a plurality of sentiments associated with the plurality of social media items, and determine a response type based on the identified sentiment of the event and supporting evidence of the determined event type; and
transmit a message associated with the determined response type to a first responder terminal; and

implementing a clustering engine configured to:
categorize the each social media item into each topic of a plurality of topics; and
in response to categorizing the each social media item into the each topic, determine whether a plurality of social media items have been categorized to a target topic within a predetermined amount of time according to evaluating time stamps of the social media items, and select the target topic when a predetermined number of social media items that have been posted to the target topic.

8. The non-transitory computer readable medium according to claim 7 , wherein the instruction code is executable by the machine for causing the machine to further implement:
word embedding logic configured to generate vector representations of each word in each of the social media items of the target topic;
a long-short term (LSTM) neural network configured generate a vector representation of predicted activities associated with the real-time data feed; and
a feed forward neural network configured to generate an output indicative of whether the social media items are associated with an event based on the vector representations of each word in each of the social media items of the target topic and the vector representation of predicted activities.

9. The non-transitory computer readable medium according to claim 8 ,
wherein the reporting engine is coupled to a knowledge graph database that defines one or more relationships between event types, sentiment types, and response types, and wherein the implementation of the reporting engine is configured to locate in the knowledge graph the response type associated with the determined sentiment and the event type and generates the recommendation based on the located response type.

10. The non-transitory computer readable medium according to claim 7 , wherein the instruction code is executable by the machine for causing the machine to implement a sentiment analyzer configured to aggregate the social media items associated with the target topic and to determine a sentiment associated with the aggregate.

11. The non-transitory computer readable mediu

Assignees

• Accenture Global Solutions Ltd

Inventors

• Limsopatham Nut
• Zaman Md Faisal
• Lecue Freddy

Classifications

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06F13/378

  using a parallel poll method · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/042 Primary

  Knowledge-based neural networks; Logical representations of neural networks · CPC title