Cryptocurrency Bubble Detection: A New Stock Market Dataset, Financial Task & Hyperbolic Models

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 5531 - 5545 July 10-15, 2022 ©2022 Association for Computational Linguistics Cryptocurrency Bubble Detection: A New Stock Market Dataset, Financial Task & Hyperbolic Models Ramit Sawhney†∗, Shivam Agarwal‡∗, Vivek Mittal†∗, Paolo Rosso△, Vikram Nanda⋆, Sudheer Chava† † Financial Services Innovation Lab, Georgia Institute of Technology, ‡University of Illinois at Urbana-Champaign, △Universitat Politècnica de València, ⋆University of Texas at Dallas shivama2@illinois.edu, {rsawhney31,schava6}@gatech.edu Abstract The rapid spread of information over social media influences quantitative trading and in- vestments. The growing popularity of spec- ulative trading of highly volatile assets such as cryptocurrencies and meme stocks presents a fresh challenge in the financial realm. In- vestigating such "bubbles" - periods of sud- den anomalous behavior of markets are critical in better understanding investor behavior and market dynamics. However, high volatility coupled with massive volumes of chaotic so- cial media texts, especially for underexplored assets like cryptocoins pose a challenge to existing methods. Taking the first step to- wards NLP for cryptocoins, we present and publicly release CryptoBubbles, a novel multi- span identification task for bubble detection, and a dataset of more than 400 cryptocoins from 9 exchanges over five years spanning over two million tweets. Further, we develop a set of sequence-to-sequence hyperbolic mod- els suited to this multi-span identification task based on the power-law dynamics of cryp- tocurrencies and user behavior on social media. We further test the effectiveness of our models under zero-shot settings on a test set of Reddit posts pertaining to 29 “meme stocks”, which see an increase in trade volume due to social media hype. Through quantitative, qualitative, and zero-shot analyses on Reddit and Twit- ter spanning cryptocoins and meme-stocks, we show the practical applicability of CryptoBub- bles and hyperbolic models. 1 Introduction Cryptocurrency (crypto) trading presents a new in- vestment opportunity (Chuen et al., 2017) for max- imizing profits. The rising ubiquity of speculative trading of cryptocurrencies over social media has lead to sentiment driven “bubbles” (Chohan, 2021; Hu et al., 2021). Such a bubble is characterized by rapid escalation of price in a short period of ∗Equal contribution. Figure 1: We present market moving tweets by influen- tial users about DogeCoin. Such tweets induce social media hype which leads to creation of bubbles. time typically driven by exuberant investor behav- ior (Fry and Cheah, 2016) and may be tied with enormous risks. Analyzing such anomalous behav- iors can be useful for forecasting speculative risks. However, it is not trivial to use existing NLP meth- ods (Sawhney et al., 2020a) for forecasting crypto bubbles as they are designed for simple assets such as equities that are 10x less volatile than crypto (Liu and Serletis, 2019). Also, crypto behavior is more strongly tied to user sentiment and social media usage as opposed to conventional stocks and equi- ties, rendering both conventional financial models and contemporary ML models weak as they are not geared towards dealing with large volumes …

Content truncated to 500 words