Dataset Collection:
- The hatespeech dataset is collected from different open sources like Kaggle ,social media like Twitter.
- The dataset has the two classes hatespeech and non hatespeech.
- The class distribution is equal
- Different strategies have been followed during the data gathering phase.
- The dataset is collected from relevant sources.
distilbert-base-uncased model is fine-tuned for Hate Speech Detection
- The model is fine-tuned on the dataset.
- This model can be used to create the labels for academic purposes or for industrial purposes.
- This model can be used for the inference purpose as well.
Data Fields:
label: 0 - it is a hate speech, 1 - not a hate speech
Application:
- This model is useful for the detection of hatespeech in the tweets.
- There are numerous situations where we have tweet data but no labels, so this approach can be used to create labels.
- You can fine-tune this model for your particular use cases.
Model Implementation
!pip install transformers[sentencepiece]
from transformers import pipeline
model_name="Sakil/distilbert_lazylearner_hatespeech_detection"
classifier = pipeline("text-classification",model=model_name)
classifier("!!! RT @mayasolovely: As a woman you shouldn't complain about cleaning up your house. & as a man you should always take the trash out...")
Github: Sakil Ansari
- Downloads last month
- 44
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.