---
license: apache-2.0
language:
- en
metrics:
- accuracy
pipeline_tag: text-classification
tags:
- Crypto
- Bitcoin
- Sentiment Analysis
- RoBERTa
- NLP
- Cryptocurrency
---

# CryptoBERTRefined
CryptoBERTRefined is a fine tuned model from [CryptoBERT by Elkulako](https://huggingface.co/ElKulako/cryptobert) model (See the base model to see it's training corpus).

# Training Process
Total of 3.803 text have been labelled manually to fine tune the model, and data augmentation is done with Back-Translation using Google Translate API with 10 language ('it', 'fr', "sv", "da", 'pt', 'id', 'pl', 'hr', "bg", "fi").

# Training Corpus
Randomly picked text from [kaggle datasets](https://www.kaggle.com/datasets/kaushiksuresh147/bitcoin-tweets)
Labelled sentiment text from [surgeAI](https://www.surgehq.ai/datasets/crypto-sentiment-dataset)

# Source Code
See [Github](https://github.com/AfterRain007/cryptobertRefined) for the source code to finetune cryptoBERT model into cryptoBERTRefined.