Automated sentiment analysis and/or geotagging of social network posts

1. A system for automatically computing a sentiment for a specific topic, comprising:

a server; and

a hardware processor of a computing device executing code for a client terminal which is in communication over a network with said server, said hardware processor executing code for:

extracting data captured by at least one sensor installed in a public place and depicting people activity in a specific geographic location corresponding to a respective topic, analyzing said extracted data to identify an indication of a sentiment value by estimating people sentiments expressed by said activity in said captured data and labelling said captured data with said indication, said at least one sensor installed in said public place is selected from a group consisting of still cameras, infrared cameras, video cameras and audio sensors;

providing to said server, an indication of the specific geographic location, the respective topic, and the data captured by the at least one sensor; and

receiving, from the server, a target sentiment value for the respective topic, wherein the server executes code for:

for each of a plurality of user generated content items uploaded to a social network from a plurality of different client devices:

selecting a specific geographic region mapping dataset of a plurality of geographic region mapping datasets corresponding to a specific geographic region identified by an analysis of the respective user generated content post,

mapping by the specific geographic region mapping dataset, a generic term to a specific geographic location within the specific geographic region,

tagging the respective user generated content item with a geo-location tag of the specific geographic location within the specific geographic region;

clustering the user generated content items according to geo-location tags and topics into a plurality of topics;

selecting at least one cluster matching the indication of the specific geographic location and the respective topic received from the client terminal;

inputting a combination of a plurality of user generated content items of the selected at least one cluster and the labelled extracted data captured by the at least one sensor received from the client terminal into a machine learning (ML) model trained on a training dataset of a combination of user generated content items and data captured by the at least one sensor and labelled with said indication of said sentiment value; and

obtaining a target sentiment value for the respective topic as an outcome of the ML model.

2. The system of claim 1 , at least one of: (i) wherein the specific topic is defined for a specific geographic location in the specific geographic region, wherein selecting comprises selecting the subset of the plurality of user generated content items having geo-location tags matching a specific spatiotemporal event, wherein extracting comprises extracting data captured by at least one sensor depicting people activity in the specific geographic location, and (ii) wherein the specific topic is a specific spatiotemporal event defined for a specific time interval in a specific geographic location in the specific geographic region, wherein selecting comprises selecting the subset of the plurality of user generated content items having geo-location tags and timestamps matching the specific spatiotemporal event, wherein extracting comprises extracting data captured by at least one sensor depicting people activity in the specific geographic location at the specific time interval.

3. The system of claim 1 , wherein said hardware processor executing additional code for:

computing a sentiment baseline as an outcome of inputting into the ML model a baseline plurality of user generated content items having a timestamp external to a specific time interval and/or geographic location external to a specific geographic location of a specific spatiotemporal event; and

analyzing the sentiment value for the specific spatiotemporal event relative to the sentiment baseline.

4. The system of claim 1 , wherein said hardware processor executing additional code for:

computing a sentiment baseline as an outcome of inputting into the ML model the baseline plurality of user generated content items having a timestamp external to a specific time interval of a specific spatiotemporal event and geographic location matching a specific geographic location of the specific spatiotemporal event; and

analyzing the sentiment value for the specific spatiotemporal event relative to the sentiment baseline.

5. The system of claim 1 , wherein said hardware processor executing additional code for:

computing a sentiment baseline as an outcome of inputting into the ML model the baseline plurality of user generated content items having a timestamp matching a specific time interval of a specific spatiotemporal event and geographic location external to a specific geographic location of the specific spatiotemporal event; and

analyzing the sentiment value for the specific spatiotemporal event relative to the sentiment baseline.

6. The system of claim 5 , wherein a first historical sentiment profile of a plurality of sentiment values computed over a historical time interval computed for the geographic location external to the geographic region of the specific spatiotemporal event has a distribution that is similar to a second historical sentiment profile of a plurality of sentiment values computed over the historical time interval for the geographic region of the specific spatiotemporal event.

7. The system of claim 1 , wherein said hardware processor executing additional code for:

computing a sentiment baseline as an outcome of inputting into the ML model a baseline plurality of user generated content items indicating content external to a specific spatiotemporal event and having a timestamp matching to a specific time interval of the specific spatiotemporal event and/or geographic location matching a specific geographic location of the specific spatiotemporal event; and

analyzing the sentiment value for the specific spatiotemporal event relative to the sentiment baseline.

8. The system of claim 1 , wherein said hardware processor executing additional code for:

clustering the subset of the plurality of user generated content items into at least two clusters according to a respective user type,

for each of the at least two clusters, computing a respective cluster sentiment value as an outcome of inputting into the ML model user generated content items of the respective cluster, thereby computing a plurality of cluster sentiment values; and

analyzing the plurality of cluster sentiment values by evaluating a first cluster sentiment value relative to a second cluster sentiment value.

9. The system of claim 1 , wherein said hardware processor executing additional code for:

identifying a plurality of users that posted the plurality of user generated content items; and

wherein performing the sentiment analysis comprises:

computing a sentiment baseline as an outcome of inputting into the ML model a baseline plurality of user generated content items comprising a plurality of historical user generated content posted by the plurality of user prior to a specific spatiotemporal event; and

analyzing the sentiment value for the specific spatiotemporal event relative to the sentiment baseline.

10. The system of claim 9 , wherein said hardware processor executing additional code for computing an increase in sentiment for the specific spatiotemporal event relative to the sentiment baseline, and/or computing a descent rate from the increase in sentiment for the specific spatiotemporal event until the sentiment baseline, the descent rate computed by iteratively computing the sentiment using corresponding sequentially posted user generated content items,

wherein the descent rate compared to the sentiment baseline indicates an impact of the spatiotemporal event.

11. The system of claim 9 , wherein the plurality of historical user generated content items match the specific geographic region of the specific spatiotemporal event prior to the specific spatiotemporal event.

12. The system of claim 11 , wherein said hardware processor executing additional code for computing an increase in sentiment for the specific spatiotemporal event relative to the sentiment baseline, and/or computing a descent rate from the increase in sentiment for the specific spatiotemporal event until the sentiment baseline, the descent rate computed by iteratively computing the sentiment using corresponding sequentially posted user generated content items, wherein the descent rate compared to the sentiment baseline indicates an impact of the spatiotemporal event.

13. The system of claim 1 , wherein said hardware processor executing additional code for:

iteratively inputting into the ML model, a historical combination of the subset of the plurality of user generated content items and the extracted data captured by the at least one sensor obtained for a historical plurality of time intervals, to obtain a plurality of historical sentiment values indicating a baseline sentiment profile,

inputting the baseline sentiment profile into a sentiment ML model that generates an outcome a predicted sentiment value for a current time interval corresponding to a time interval of the sentiment value; and

analyzing the sentiment value relative to the predicted sentiment value.