ritvik-garg/hate-speech-detection

Multi-Lingual Sexist Hate Speech Detection and Classification

sEXism Identification in Social neTworks (EXIST) Dataset

Dataset has been developed to address this critical social problem. The problem can be divided into two tasks:

• The first task, Sexism Identification, is a binary classification problem to classify the text as sexist or non-sexist.
• The second task, Sexism Categorization, is a multi-class classification problem where texts labelled as sexist are classified into one of five labels: ideological-inequality, stereotypingdominance, objectification, sexual-violence and misogyny-non-sexual-violence

Our Project is divided into 8 files. Following are the files with the description of what it contains:

• XLM_R.ipynb: Contains XLM-R model code and results
• ROBERTA.ipynb: Contains Roberta model code and results
• Multilinguial_BERT.ipynb: Contains Multilinguial BERT model code and results
• BERT_and_BETO.ipynb : Contains BERT and BETO models codes and results
• Bertweet.ipynb: Contains Bertweet model code and results
• Exist_Data_Augmentation.ipynb: Contains code for data augmentation and stored results on drive
• Augmented_BERT_&_BETO.ipynb: Contains BERT and BETO models trained on augmented data codes and results
• Class_Balanced_BERT_&_BETO.ipynb: Contains BERT and BETO models trained on class balanced data codes and results

Steps to Run

• Create a folder named SEXISM DATA on your google drive
• Upload the 2 train and test files in this folder
• Now run each of the above 8 code files independently from colab.
• Run all cells of each file in sequence.
• This will fetch the results.

Results

All results are in the report as well as the individual results for each model are in their respective python notebooks file also.

Project Home Page

Link : http://nlp.uned.es/exist2021/