Multilingual Detection of Hate Speech Against Immigrants and Women in Twitter at SemEval-2019 Task 5: Frequency Analysis Interpolation for Hate in Speech Detection

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 460–463 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 460 OscarGaribo at SemEval-2019 Task 5: Frequency Analysis Interpolation for Hate in Speech Detection `Oscar Garibo i Orts Universitat Polit`ecnica de Val`encia / 46025 Val`encia Spain osgaor@alumni.upv.es Abstract This document describes a text change of rep- resentation approach to the task of Multilin- gual Detection of Hate Speech Against Im- migrants and Women in Twitter, as part of SemEval-20191. The task is divided in two sub-tasks. Sub-task A consists in classifying tweets as being hateful or not hateful, whereas sub-task B requires fine tuning the classifica- tion by classifying the hateful tweets as being directed to single individuals or generic, if the tweet is aggressive or not. Our approach con- sists of a change of the space of representation of text into statistical descriptors which char- acterize the text. In addition, dimensional re- duction is performed to 6 characteristics per class in order to make the method suitable for a Big Data environment. Frequency Analy- sis Interpolation (FAI) is the approach we use to achieve rank 5th in Spanish language and 9th in English language in sub-task B in both cases. 1 Introduction Social media has become a new standard of com- munications in the last years. Every year more and more people actively participate in the content cre- ation, sometimes under the shield of anonymity. Social media has become a complex communica- tion channel in which usually offensive contents are written. Supervising the content and banning offensive messages currently is a subject of high interest for social media administrators. Offensive speech can be addressed to individuals or groups due to the race, sexuality, religion and some other characteristics. In this task two of these character- istics will be used as target for offensive speech, women and immigrants. This problem will be con- sidered as an Author Profiling task, since the main 1alt.qcri.org/semeval2019/ goal is building a system which would ideally de- tect author whose content is offensive to women and/or immigrant. Author Profiling is widely stud- ied and some new ideas arise from time to time (Rangel et al., 2016). We have developed a new representation method for text that reduces the di- mensionality of the information for each author to 6 characteristics per class. This representation, Frequency Analysis Interpolation, is used to cod- ify the texts for each user and this codified infor- mation is used as input data to support vector ma- chines with linear kernel. In a Big Data environ- ment, reducing the number of characteristics from thousands to 6 per class allows an efficient way to deal with high volumes at high speed. With this will in mind a previous method was tested which can be checked at (Garibo, 2018). 2 Corpus Two corpora have been created to be used in Se- mEval Task5, HatEval (Basile et al., 2019). One in each of the 2 different languages which are sub- ject of …

Content truncated to 500 words