الفهرس 
Generic Event DetectionOnly 14 pages are availabe for public view
 |  | from | 122 |  |  |
| | | from | 122 | | |
Abstract
Every day, millions of people write and trade news on social media, making it a major source of information. Therefore, the use of social media in everyday life has become a necessity for keeping up with the news, as well as making inquiries and requesting assistance. When an event occurs, news spread quicker on social media than on other news sites, making them a good source for event detection. Text analysis is the most popular method for detecting events in social networks.
In the event of an emergency, social media can be quite useful in acquiring a better knowledge of the situation. The information about the situation starts to circulate on social media, with the aim of raising awareness and ensuring that everyone is aware of all important instructions and can request help. This occurs due to the fact that this information is accessed directly from those who are affected. If the information gathered is successfully used, it may be used to respond to people with the appropriate needs.
Trying to reach the affected people to help them through social media is a difficult process in the presence of a lot of data circulating on these sites, which require appropriate techniques to extract the required information during the occurrence of the crisis. In this work, we proposed a hybrid approach for detecting affected people and their needs during crises that is based on combination of text analysis techniques and location identification process.
We proposed a hybrid strategy for finding impacted people during crises and extracting their needs in terms of asking for help that incorporates text analysis and location identification algorithms. The use of location data is done to filter out persons who are writing about the situation without being affected. The experiment on Twitter data revealed that combining text analysis with location produced better results, with an accuracy of 96 % against 87 % when using text analysis alone.
We tried to detect the affected people’s needs and answer them with the suitable instructions and guidelines by using question-answering techniques. These techniques are based on natural language processing techniques and neural networks to extract the needs of those who have been impacted and respond appropriately. The proposed approach provide appropriate guidance with a precision of 0.81, a recall of 0.76 and an f-score of 0.78. we testing our approach using twitter data from various type of crises.