Fine-tuned Seq2Seq model is developed for analysing and summarization of cryptocurrency news for the following crypto coins:
Bitcoin, Ethereum, Tether, Solana, Binance Coin. Max input size for texts is 1024 tokens that is about
3.5K chars of texts. Model is created by fine-tuning facebook/bart-large transformer model.
Model outputs short text summary and uptrend/downtrend lists of specified above crypto coins if their trends are considered in the news text.
1summarizer = pipeline("summarization", model = "bpavlsh/bart-crypto-summary")
2txt="""
3Crypto market shows mixed signals. Bitcoin (BTC) and Ethereum (ETH) is experiencing a slight downturn, weighed down by bearish
4investor sentiment, while Solana (SOL) see sharp uptrends driven by increased on-chain activity.
5"""
6result=summarizer(txt, early_stopping=True)[0]['summary_text']
7print(result)
8
9Result:
10"""
11Bitcoin and Ethereum are experiencing a slight downturn with bearish investor sentiment, while Solana shows a strong uptrend driven by increased on-chain activity.
12Uptrend: Solana.
13Downtrend: Bitcoin, Ethereum.
14"""
We are sharing a considered model and results for academic purpose only,
not any financial advice or recommendations for real business or investment.
Pavlyshenko B.M. Financial News Analytics Using Fine-Tuned Llama 2 GPT Model. arXiv preprint arXiv:2308.13032. 2023. Download PDF:
https://arxiv.org/pdf/2308.13032.pdfPavlyshenko B.M. Analysis of Disinformation and Fake News Detection Using Fine-Tuned Large Language Model. arXiv preprint arXiv:2309.04704. 2023. Download PDF:
https://arxiv.org/pdf/2309.04704.pdfPavlyshenko, B.M. Bitcoin Price Predictive Modeling Using Expert Correction. 2019 XIth International Scientific and Practical Conference on Electronics and Information Technologies (ELIT), September 16 – 18, 2019 Lviv, Ukraine, pages: 163-167. Download PDF:
https://arxiv.org/pdf/2201.02729
