From: How are texts analyzed in blockchain research? A systematic literature review
Paper | Citation | CPY | Data and period | Methodology | Summary |
---|---|---|---|---|---|
Polasik et al. (2015) | 431 | 54 | Nexis database: English-language news mentioning Bitcoin 04.2011– 03.2014 | Lexicon-based sentiment analysis (Henry’s finance-specific dictionary) | This study examines the determinants of Bitcoin price and the drivers of its success. Using sentiment in newspaper articles as one of the variables, it discovers that the negative mentions of Bitcoin lead to a price drop, while exhortatory pieces increase the Bitcoin price. |
Kim et al. (2016) | 268 | 38 | Comments and relevant replies in three cryptocurrency online communities: Bitcointalk, Forum ethereum, Xrpchat 12.2013–02.2016 | Lexicon-based sentiment analysis (VADER) | This study uses the contents on three cryptocurrency communities to predict the price and number of transaction fluctuations. The sentiment of posts, the number of posts/replies, and the number of views of posts are used to perform Granger causality test on each currency for a time lag of 1–13 days. The results show that positive comments affect the price fluctuations of Bitcoin, whereas Ethereum and Ripple are influenced by negative comments. |
Mai et al. (2018) | 208 | 42 | Bitcointalk 01.2012–12.2014 Twitter: hastag Bitcoin 09.2014–12.2014 | Lexicon-based sentiment analysis (LM lexicon) | This study investigates the impacts of social media on Bitcoin price. It separates the users into two groups, (1) the silent majority of users and (2) the vocal minority, and examines the impacts of these two groups, respectively. It finds that Bitcointalk has a more substantial impact than Twitter, and the silent minority exerts a more significant effect on future Bitcoin prices. |
Georgoula et al. (2015) | 170 | 21 | Twitter: keywords and hashtags Bitcoin, BTC, and Bitcoins 10.2014–01.2015 | Machine leaning-based sentiment analysis (Support Vector Machines) | This study sheds light on the factors determining the price of Bitcoin in the short- and long-run. It adds Twitter sentiment into conventional prediction model. Specifically, it constructs a Twitter sentiment measure using SVMs and finds that sentiments have a positive short-run impact on Bitcoin prices. |
Abraham et al. (2018) | 164 | 33 | Twitter: hashtags Bitcoin and Ethereum 03.2018–05.2018 | Lexicon-based sentiment analysis (VADER) | This study uses a linear model for predicting price changes of Bitcoin and Ethereum utilizing Twitter sentiment, tweet volume and Google Trends data. The results indicate that Twitter sentiment tends to be positive regardless of price direction and is, therefore, not a feasible predictor of price changes. |
Kraaijeveld and de Smedt (2020) | 132 | 44 | Twitter: hashtags including following nine cryptocurrencies: Bitcoin, Ethereum, XRP, Bitcoin Cash, EOS, Litecoin, Cardano, Stellar and TRON 06.2018–08.2018 | Lexicon-based sentiment analysis (VADER, LM lexicon, and manually complied cryptocurrency-related words) | This study tests to what extent Twitter sentiment can be used to predict price returns for nine cryptocurrencies. It measures sentiments using a self-constructed lexicon and performs bilateral Granger-causality testing to find the causality. It finds the predictive power of Twitter sentiment for several cryptocurrencies. |
Grover et al. (2019) | 122 | 31 | Twitter: hashtag Blockchain 01.2018–02.2018 | Lexicon-based sentiment analysis (Bing) | This study explores blockchain acceptance by examining the tweet information. It combines manual content analysis and lexicon-based sentiment analysis to distinguish the topics discussed and the user opinion. The analysis shows that users are attracted by security, privacy, transparency, trust and traceability. Furthermore, blockchain benefits are more frequently discussed than its drawback. |
Karalevicius et al. (2018) | 120 | 24 | Expert media news from CoinDesk, Cointelegraph, NewsBTC 05.2013–02.2016 | Lexicon-based sentiment analysis (Harvard-IV General Purpose Psychological Dictionary and LM lexicon) | This study utilizes Bitcoin-related news articles to predict semi-short-term Bitcoin price movement. Integrating the sentiments of such news shows that the market initially overreacted to the news articles, resulting in multiple corrections. |
Li et al. (2019) | 93 | 23 | Twitter: keywords and hashtags ZClassic, ZCL, and BTCP 01.2019–02.2019 | Lexicon-based sentiment analysis (Textblob) | This study analyzes Twitter signals as a medium for user sentiment to predict the hourly price fluctuations of ZClassic. It compiles the tweets into an hourly sentiment index, creating a weighted index giving larger weight to retweets. These two indices and the raw sentiment are used as input for Extreme Gradient Boosting Regression Tree Model for prediction. |
Valencia et al. (2019) | 90 | 23 | Twitter: keywords and hashtags including following four cryptocurrencies: Bitcoin, Ethereum, XRP, Litecoin 02.2018–04.2018 | Lexicon-based sentiment analysis (VADER) | This study uses sentiments on Twitter as input features for multiple machine learning algorithms to predict the price movement of four cryptocurrencies. It shows that Twitter data alone can be used to predict certain cryptocurrencies. |
Kim et al. (2020) | 79 | 26 | Academic papers: keyword or abstract contain “Blockchain”, “Block chain”, and “Block-chain” in six databases: Scopus, ScienceDirect, Web of Science, IEEE Xplore, Google Scholar, and Korean Citation Index. 01.2014 - 08.2018 | Topic modeling (W2V-LSA) | This study proposes an improved method for topic modeling (W2V-LSA) and performs an annual trend analysis of blockchain-related literature. The experimental results confirmed the usefulness of W2V-LSA in terms of the accuracy and diversity of topics by quantitative and qualitative evaluation, and it can be an option for researchers using topic modeling for technology trend analysis. |