Compilación y etiquetado de corpus para el análisis de la violencia lingüística en Twitter: problemas y soluciones

Identificador:  imarina:9443071

Handle: https://hdl.handle.net/20.500.11797/imarina9443071

Autors:  Susana María Campillo Muñoz; María Dolores Jiménez-López

Resum:
The interest in the study of verbal violence is increasing in linguistics and computational branches. However, investigators try to face some issues, such as conceptual definitions, and the categories included in the analysis. In computational analysis specially, there are problems with the annotation task and the linguistic features extracted. In order to detect the problems in verbal violence corpora analysis and to propose some solutions, we simulate a tagging task with a sample corpus of 100 tweets between three annotators. They must tag every tweet as violent or non-violent. Our results confirm these differences in the comprehension of verbal violence and some problems related to hashtags and emojis in the computational analysis. Then, we propose some solutions related to the annotation task and the computational analysis. With the aim of getting a common concept of verbal violence in the annotation task, we need to use an annotation scheme. Also, it is necessary to create a list of different linguistic features, from emojis to situational attributes, for improving the computational analysis. To sum up, linguistics and computation need to work together so that we could achieve best results in the analysis of verbal violence.