Paper: arXiv 2402.10481

Authors: Xiaorui Zuo, Yao-Tsung Chen, Wolfgang Karl Härdle

Abstract

In the burgeoning realm of cryptocurrency, social media platforms like Twitter have become pivotal in influencing market trends and investor sentiments. In our study, we leverage GPT-4 and a fine-tuned transformer-based BERT model for a multimodal sentiment analysis, focusing on the impact of emoji sentiment on cryptocurrency markets. By translating emojis into quantifiable sentiment data, we correlate these insights with key market indicators like BTC Price and the VCRIX index. Our architecture’s analysis of emoji sentiment demonstrated a distinct advantage over FinBERT’s pure text sentiment analysis in such predicting power. This approach may be fed into the development of trading strategies aimed at utilizing social media elements to identify and forecast market trends. Crucially, our findings suggest that strategies based on emoji sentiment can facilitate the avoidance of significant market downturns and contribute to the stabilization of returns. This research underscores the practical benefits of integrating advanced AI-driven analyses into financial strategies, offering a nuanced perspective on the interplay between digital communication and market dynamics in an academic context.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.0)

Why this score: The paper employs advanced deep learning architectures (GPT-4 and fine-tuned transformer models) and statistical correlation with market indices, indicating moderate-to-high mathematical sophistication. It also demonstrates strong empirical rigor with a concrete dataset of social media text, specific market indicators (BTC Price, VCRIX), and a defined backtest-ready trading strategy based on sentiment signals.

Research Flowchart

  flowchart TD
  A["Research Goal: Analyze Emoji Impact on Crypto Markets"] --> B["Data Inputs: Twitter Data"]
  B --> C["Multimodal Processing: GPT-4 & Fine-tuned BERT"]
  C --> D["Emoji Quantification & Correlation"]
  D --> E["Compare vs. FinBERT Text Analysis"]
  E --> F["Key Findings: Emoji sentiment outperforms text"]
  F --> G["Outcomes: Predict Market Trends & Avoid Downturns"]
  style A fill:#e1f5fe
  style G fill:#e8f5e8

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a multimodal framework for emoji sentiment analysis and example trading strategies based on that sentiment.

Source sentence
In our study, we introduced a novel multimodal framework for unsupervised emoji sentiment analysis of cryptocurrency-related tweets, leveraging the advanced capabilities of GPT-4 and Bert.

Assumes 5

  • Sentiment derived from emojis can be quantified and used as a predictor of market variables. model
    Source sentence
    By translating emojis into quantifiable sentiment data, we correlate these insights with key market indicators like BTC Price and the VCRIX index.
  • The selected BTC price window aligns with the Twitter sentiment window for proper correlation. data
    Source sentence
    We have selected the Bitcoin (BTC) closing price data for the period starting from the 8th of March, 2019, through to the 23rd of November, 2019. This specific interval was chosen to ensure a robust alignment with the timeline of the Twitter Emoji sentiment analysis.
  • Trading strategies are evaluated without accounting for transaction costs. execution
    Source sentence
    In this paper, however , the proposed trading strategies do not account for transaction costs, an aspect that warrants consideration in real‑world applications.
  • Tweets containing the keyword "crypto" and emojis represent a diverse and unbiased sample of market sentiment. data
    Source sentence
    Our training dataset comprises a collection of tweets harvested from Twitter, utilizing the Twitter API. The focus of this dataset is on tweets post by the accounts containing the keyword "crypto" and incorporating emojis.
Show 1 more
  • The sentiment average of top‑n daily emoji sentiments captures the positive market signal. statistical
    Source sentence
    Our research reveals a significant positive correlation between daily top 5 and top 10 tweets ’ average emoji sentiments() and the subsequent day's Bitcoin (BTC) prices.

Data

  • Twitter training dataset · Twitter API text-news public
    2023-11-08 to 2024-01-28 · tweets containing "crypto" and emojis
    Source sentence
    Our training dataset comprises a collection of tweets harvested from Twitter, utilizing the Twitter API
  • Bitcoin tweets - 16M tweets With Sentiment Tagged · Kaggle text-news public
    Bitcoin‑related tweets filtered for emojis
    Source sentence
    Our test dataset originates from the "Bitcoin tweets - 16M tweets With Sentiment Tagged" dataset available on Kaggle
  • BTC closing price data market-prices public
    2019-03-08 to 2019-11-23 · Bitcoin (BTC)
    Source sentence
    We have selected the Bitcoin (BTC) closing price data for the period starting from the 8th of March, 2019, through to the 23rd of November, 2019.
  • VCRIX volatility index · Kim 2021 market-prices public
    cryptocurrency market volatility
    Source sentence
    VCRIX index captures the volatility changing of VCRIX (Kim, A. 2021) index

Recognised sources: Twitter / X Kaggle

Costs

Not modeled A trading or execution result is reported without costs.

Source sentence
In this paper, however , the proposed trading strategies do not account for transaction costs, an aspect that warrants consideration in real‑world applications.

Finds 3

  • Daily top‑5 and top‑10 emoji sentiment averages positively predict the next day Bitcoin price, while bottom sentiment shows no significant relationship. predictive-accuracy
    Source sentence
    Our research reveals a significant positive correlation between daily top 5 and top 10 tweets ’ average emoji sentiments() and the subsequent day's Bitcoin (BTC) prices, while the bottom 5 and bottom 10 tweets’ average emoji sentiments() each day show no significant relationship.
  • Emoji sentiment also exhibits predictive power for the weekly trend of the VCRIX volatility index. predictive-accuracy
    Source sentence
    Our findings reveal that emoji sentiment exhibits a significant predictive power on subsequent day Bitcoin prices and the weekly trend of the VCRIX volatility index.
  • The proposed trading strategies generate consistent positive returns across various time windows and parameter values. performance
    Source sentence
    The strategies demonstrated consistent positive returns across various time windows and values of , particularly highlighting their effectiveness in loss aversion and stabilizing profits.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 22 related papers). Coverage is partial, so this is neither replication nor refutation.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (27)

  1. Are markets sentiment driving the price bubbles in the virtual? (2024) — link
  2. Sentiment Classification of Cryptocurrency-Related Social Media Posts (2023) — link
  3. Hedging cryptos with Bitcoin futures (2023) — link
  4. GPT-4 Technical Report (2023) — link
  5. Hedging cryptocurrency options (2021) — link
  6. Media-expressed Tone, Option Characteristics, and Stock Return Predictability (2021) — link
  7. Integrating online and offline data for crisis management: Online geolocalized emotion, policy response, and local mobility during the COVID crisis (2021) — link
  8. Deep learning-based cryptocurrency sentiment construction (2020) — link
and 19 more
  1. Pricing Cryptocurrency Options (2020) — link
  2. VCRIX — A volatility index for crypto-currencies (2019) — link
  3. CRIX an Index for cryptocurrencies (2018) — link
  4. Singularity-constrained octahedral fields for hexahedral meshing (2018) — link
  5. Twitter Sentiment Analysis via Bi-sense Emoji Embedding and Attention-based LSTM (2018) — link
  6. Using millions of emoji occurrences to learn any-domain representations for detecting sentiment, emotion and sarcasm (2017) — link
  7. Joint Emoji Classification and Embedding Learning (2017) — link
  8. emoji2vec: Learning Emoji Representations from their Description (2016) — link
  9. Distillation of News Flow Into Analysis of Stock Reactions (2016) — link
  10. CRIX or evaluating Blockchain based currencies (2015) — link
  11. News or Noise? Using Twitter to Identify and Understand Company‐specific News Flow (2014) — link
  12. The VIX, the Variance Premium and Stock Market Volatility (2013) — link
  13. Exploiting emoticons in sentiment analysis (2013) — link
  14. MoodLens: an emoticon-based sentiment analysis system for chinese tweets (2012) — link
  15. Emoticon Smoothed Language Models for Twitter Sentiment Analysis (2012) — link
  16. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (2019) — link
  17. Media-Expressed Tone, Option Characteristics, and Stock Return Predictability (2019) — link
  18. Alleviating Data Sparsity for Twitter Sentiment Analysis (2012)
  19. ZUO CHE Hae Emoji Sentiments reactions

0 of 27 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.