Where are you posting from?

Where are you posting from?

Problem being addressed

Over the last decade​,​ social media has become a prominent channel for information exchange and opinion communication. While social media platforms allow their users to tag messages with geographic tags in the form of longitude and latitude coordinates​,​ understanding the semantics of these locations​,​ e.g. whether messages were posted from a restaurant​,​ a museum or a church​,​ remains complicated.

Solution

Automatic location type classifier that can learn the location type based on the content of the messages sent from this location as well as their context information (e.g. their sending time). To explore this task the researchers have collected a dataset of locations – where each location is defined as a certain radius around a given (longitude, latitude) coordinates– and geo-tagged tweets sent from a close geographical proximity to each location. Each location is annotated with its location type: school, university, church, shop, museum or a health location.

Advantages of this solution

The tests show high accuracy of the suggested joint approach and allows the researchers to further explore contextual features as well as more elaborated linguistic features, and to design structure-aware models that properly account for the inter-message relations in the dataset.

Solution originally applied in these industries

media

Media

Possible New Application of the Work

/static/media/childcare-1.b132e1e8.png

Childcare Sector

The importance of the location classifier task goes beyond the opinions and information posted in social media. Consider, for example, the detection of unauthorized businesses such as day-cares. By classifying residential buildings as education-related, based on the posts authored at these buildings, illegal day-care providers can be detected, such places can risk their attendees because they are not supervised by the authorities.

/static/media/others.cca5db97.png

Other Industry

Security: proper location classifier allows timely reactions to disasters or events like terrorist attacks or mass shootings, helping the police identify the exact place from the tweets.

Author of original research described in this blitzcard: Elad Kravi, Benny Kimelfeld, Yaron Kanza, Roi Reichart

question-mark-icon

Name of the author who conducted the original research that this blitzcard is based on.

Source URL: #############check-icon


search-iconBrowse all blitzcards