Leveraging Events Sub-Categories for Violent-Events Detection in Social Media
| Type of publication: | Conference paper |
| Citation: | Vallejo-Aldana_IBERLEF2022_2022 |
| Publication status: | Published |
| Booktitle: | Proceedings of the Iberian Languages Evaluation Forum (IberLEF 2022) |
| Volume: | 3202 |
| Year: | 2022 |
| Month: | September |
| URL: | http://ceur-ws.org/Vol-3202/da... |
| Abstract: | This paper describes our participation in the shared evaluation campaign of DA-VINCIS@IberLEF 2022. In this work, we addressed the Violent Event Identification (VEI) task by exploiting Bidirectional Encoder Representations from Transformers (BERT) in combination with Multi-Task learning approaches. Our results indicate that the proposed architecture is able to leverage information about the crime categories for effectively detect the mention of a violent act within a tweet. Our approach obtained the best performance (F1=0.77) among 11 different teams and a total of 32 different submissions. |
| Keywords: | |
| Projects |
Idiap |
| Authors | |
| Added by: | [UNK] |
| Total mark: | 0 |
|
Attachments
|
|
|
|
|
Notes
|
|
|
|
|
