Hello, I would like to create a developer tool to track/monitor hate speech detection on Twitter. I would like to create a classification model that can detect hate speech in tweets using the tweet-eval dataset, but after completing the third chapter of the course I’m having trouble. Can someone lead me to resources clarifying how I can fine-tune a speech detection model using the “bert-base-uncased” as my checkpoint?
hi @BinaryCoffee .
i think you can use BertForSequenceClassification finetuning to ust hate speech detection.
the process of fine-tuning is
- preprocessing custom dataset for fine tuning
- build data loader to load custom dataset
- build fine-tuning model (hugging-face support many library of fine_tuning)
- load pre-trained model and fine-tuning
- metric performance fine-tuned model.
there are some helpful link to you.
hugging face docs of BertForSequenceClassification.
[BertForSequenceClassification docs]
thank to HF team and @nielsr !
they support various fine-tuning tutorial of various models with colab.
[huggingface transformer tutorial page]
hope to help.
regards.
2 Likes