In this research, we examined cryptocurrency data, concentrating on a specific group of cryptocurrencies. Our choice of these particular coins was driven by their significant popularity among users, as well as the limited availability of substantial data for other coins. To interpret the data, we applied four analytical methods explained in section"Introduction". Here we present the outcomes of our analysis for each of the aforementioned cryptocurrencies. The selection of features was made considering their past influence29,61. In the analysis conducted, LIWC assessments were applied to nine cryptocurrencies, resulting in an extensive collection of nine distinct analyses. We selected values that were highly informative for extracting linguistic interpretations relevant to cryptocurrencies. Our choice was made to capture key aspects of sentiment, linguistic style, and thematic content pertinent to discussions around cryptocurrencies. By narrowing down our focus to these particular features, we aimed to mine information from the psychological and linguistic dimensions of cryptocurrency discourse, thus aligning analysis with our goals. these categories encompass analytical thinking (metric of logical, formal thinking), clout (language of leadership), drives (related to personal motivations and psychological desires), affect (linguistic expressions associated with emotional and affective states expressed by a given text), money (refers to a set of linguistic cues or indicators related to financial terms, wealth, and economic aspects, Want (a human ability that allows individuals to envision future events with flexibility), attention (crucial subset of the Perception category), netspeak (represents a subset of the conversational category) and filler (non-essential sounds, words, or phrases, commonly used in speech to fill in pauses and maintain the flow of conversation without altering its meaning). In the drives and affect categories, additional features will be elaborated upon in the following discussion. Our examination indicated that Fantom attracts a larger number of tweets centered on technical aspects and holds a higher level of trust in comparison to other cryptocurrencies. For Binance, our observations revealed that the tweets predominantly revolve around themes of affiliation, achievements, and the pursuit of power and wealth. This pattern in discussions on Binance suggests a focus on notable accomplishments and financial success, indicative of a unique narrative and sentiment surrounding the coin. For Matic, the tweets primarily center around emotional impact compared to other cryptocurrencies. This emphasis on affective responses suggests that the coin is particularly influenced by emotional novelty. This distinctive characteristic could be considered a contributing factor to the fluctuations in the coins price, as emotional sentiment plays a significant role in shaping market dynamics and investor behavior. Our analysis revealed that Dogecoin exhibits a higher prevalence of netspeak, the informal language commonly used on the internet, compared to other cryptocurrencies. Conversely, Ethereum appears to attract more attention relative to other coins. This distinction suggests that Dogecoin is characterized by a more casual and internet-centric communication style, while Ethereum stands out for its ability to capture increased Attention and interest. A deeper understanding of the communication dynamics and community sentiment surrounding different coins may aid investors in making more informed choices, aligning their investment strategies with the unique qualities and trends associated with each cryptocurrency. From an emotional perspective, most cryptocurrencies exhibit a generally moderate and harmonious emotional profile. Notably, there is a distinct focus on the emotional category of Anticipation, with Dogecoin taking the forefront in this aspect. In this context, Anticipation likely signifies the expectation or excitement surrounding the future prospects, developments, or events associated with these cryptocurrencies.The outcomes of our analysis are presented in Table5. In terms of readability, the analysis revealed that Dogecoins tweets are relatively more challenging to read and comprehend, as indicated by lower scores on the Flesch Reading Ease measure. The Flesch-Kincaid and Dale-Chall Measures suggest an average reading difficulty level akin to content tailored for college graduates. Conversely, Ethereums tweets, as per the Gunning Fog Index, demand a higher level of reading proficiency, indicating a more complex and advanced readability suitable for individuals with a college-level education and vocabulary. To explore additional results, refer to Figs. 5 and 6s, as well as Table6.
The LIWC model revolutionized psychological research by making the analysis of language data more robust, accessible, and scientifically rigorous than ever before. LIWC-22 examines over 100 textual dimensions, all of which have undergone validation by esteemed research institutions globally. With over 20,000 scientific publications utilizing LIWC, it has become a widely recognized and trusted tool in the field62 giving way to novel approaches in analysis63,64. Although LIWC provides several benefits, it has its limitations. One drawback is its dependence on predefined linguistic categories, which might not encompass nuances and variations present in natural language. Furthermore, LIWC may encounter challenges in accurately deciphering sarcasm, irony, and other subtle forms of language usage, potentially resulting in text misinterpretation.
To effectively convey the outcomes of our analysis, average values among all the tweets were computed for each of LIWC categories. Averages can help identify broadscale sentiment trends over time. By tracking changes in average scores across key linguistic categories, such as sentiment, emotion, or cognitive processes, one can observe shifts in user sentiment and attitudes towards cryptocurrencies, market developments, or external events. Therefore, the average was calculated by summing up the scores of all comments related to each coin for each LIWC feature and then dividing by the total number of comments for that coin. These computed averages provide information along the linguistic and psychological dimensions intertwined with the selected digital currencies. A comprehensive presentation of these average values for each category can be found in Table3.
Analytical Thinking, when showing high numerical values, signifies a formal, logical, and hierarchical thought process. Conversely, lower numbers suggest a more informal, personal, present-focused, and narrative style of thinking65. The values of this category computed for tweets related to cryptocurrency, reach their highest average score of 67.76 in texts mentioning Fantom. This fact indicates that, on average, discussions in this domain exhibit a relatively high level of logical and formal thinking. Conversely, the lowest average score of 52.00 was found for Ripple, which might suggest that discussions concerning this particular cryptocurrency place slightly less emphasis on logical and analytical thinking compared to the cryptocurrency domains average.
Clout is one of the four summary variables in LIWC designed to assess the degree of confidence and certainty conveyed in the text66,67. Our analysis revealed that the cryptocurrency Fantom exhibits a relatively high Clout score, with an average result of 70.91. This suggests that discussions and conversations related to Fantom often convey a strong sense of confidence and certainty. This high Clout score may also indicate a substantial degree of assurance in Fantom stability. In contrast, the cryptocurrency Ripple demonstrates a comparatively lower Clout score with an average result of 43.39. Figure2 presents a comparative evaluation of Analytical Thinking and Clout scores across different cryptocurrencies. This suggests that discussions related to Ripple may not consistently display the same level of confidence and certainty found in the Fantom discussions. In essence, when Fantom demonstrates higher Clout values, it signifies that the users who composed the tweets are expressing increased confidence. This, in turn, leads us to infer a heightened level of knowledge on their part. In both analyses, we observed that Fantom consistently had the highest scores, indicating a higher level of analytical thinking and confidence in discussions related to it. Conversely, Ripple consistently had the lowest scores in both categories, suggesting a relatively lower emphasis on analytical thinking and a lower degree of expressed confidence in discussions related to it. While these observations suggest a correlation between analytical thinking and confidence in these specific cryptocurrency discussions, its important to note that correlation does not imply causation. Other factors, such as market conditions, community sentiment, and news events, can also influence these results. For example, when we examined Binance, we foound that it ranks as the second-highest in terms of Analytical Thinking scores among the various cryptocurrencies. However, when we assess it as the position in the Clout category, Binance ranks fifth. The results of Analytical Thinking and Clout analysis related to digital currencies can be viewed in Table3.
Comparative evaluation of analytical thinking and clout scores across different cryptocurrencies.
Drives is a comprehensive dimension that encapsulates various needs and motives65. In our LIWC analysis, we concentrated on the Drives, particularly examining the aspects of Affiliation, Achievement, and Power. We observed that the presence of Affiliation-related language (such as us and help) is comparatively lower in discussions related to Cardano, while it appears more frequently in conversations about Dogecoin. Similarly, in terms of Achievement-related language (including work, better, and best), Dogecoin tends to have fewer instances compared to Matic. Furthermore, when examining Power-related language (like allow and power), we found that Dogecoin exhibits a lower frequency, while Bitcoin discussions tend to feature a greater occurrence of such language. These patterns highlight variations in linguistic expressions across different cryptocurrencies, shedding light on the distinctive characteristics of discussions over different digital coins. Upon closer examination, it became evident that tweets originating from Binance sources tended to include a higher frequency of words associated with Drives, whereas Fantom source tweets had a notably lower occurrence of Drives-related words.Additional details can be found in Fig.3.
Frequency of language associated with affiliation, achievement, power, and drives across different cryptocurrency discussions.
In the Affect1 subset, our analysis encompassed various emotional dimensions, including Positive Emotion, Negative Emotion, Anxiety, Anger, Sadness, and Swear Words. In the upcoming Emotion section, we delve deeper into affective analysis. However, in this preliminary report, we provide an overview of the affective processes observed in the LIWC analysis. It can be observed in Table3 that there is a variation in affective (good, well, new, love) content among different cryptocurrencies. Notably, Matic coin exhibits a higher level of affective language, while Ada appears to have a lower level. This distinction becomes clearer when we explore the affective subcategories including Positive tone (new, love), Negative tone (bad, wrong, too much, hate), Emotion (good, love, happy, hope), and Swear words (shit, fuckin*, fuck, damn), as depicted in Fig. 4. It becomes evident that Matic coin scores higher in Positive tone and Emotion, while Bitcoin registers a higher Negative tone. Additionally, Ripple stands out with a higher score in Swear words, indicating potential user dissatisfaction. When we further break down the Emotion category into its subsets, which encompass Anxiety (worry, fear, afraid, nervous), Anger (hate, mad, angry, frustr), and Sadness (sad, disappoint, cry), we notice that Dogecoin exhibits a higher score in Anxiety, Ripple in Anger, while most of the nine analyzed coins show similar values for Sadness. These observations contribute to our analysis and highlight the varying affective language usage across different cryptocurrencies, which we explore in greater detail in the subsequent Emotion section.
Comparative analysis of affective language dimensions-positive tone, negative tone, emotion, and swear words-across different cryptocurrencies.
Want words signify the authors desires or preferences. Typically, wants are philosophically differentiated from needs by conceptualizing needs as innate and essential for survival, while wants are learned and generally linked to additional satisfaction beyond basic necessities68. What is important for cryptocurrency analysis in this category is the aspect of hope (want, hope, wanted, wish) as Want, or Hope, is a remarkable human ability that allows individuals to envision future events and their potential outcomes with flexibility69. Many users have high hopes for the future of cryptocurrency, anticipating greater benefits from their investments. From Table3, it becomes evident that Shiba is the cryptocurrency that garners most hope among users. The range of hope scores falls between 0.19 and 0.41, with the lowest level of hope associated with Fantom. This data suggests that Shiba is particularly promising in the eyes of cryptocurrency enthusiasts, while Fantom elicits comparatively less optimism.
Another important LIWC category is Money (business, pay, price, and market)22. The range of Money scores, from 2.46 for Shiba to 10.51 for Binance, indicates varying degrees of discussion or emphasis on cryptocurrency financial aspects. Notably, Binance stands out with the highest score, suggesting a significant emphasis on business and financial aspects in discussions related to this coin. Conversely, Shiba has the lowest score, indicating relatively less emphasis on these financial terms in conversations related to it. These findings offer a glimpse into the importance placed on financial and business-related aspects and potentially shed light on the perception and use of the cryptocurrencies in the broader context of market and economy.
At the dawn of experimental psychology, William James wrote that everyone knows what attention is. It is the taking possession by the mind, in a clear and vivid manner70. When users include the term Attention in their tweets, it signifies their intention to draw focus to a significant event or topic. Upon reviewing Table3, it becomes evident that Ethereum tweets receive more attention than tweets about the other cryptocurrencies, indicating a heightened interest or emphasis on Ethereum-related matters. Conversely, tweets concerning Dogecoin appear to attract less attention when compared to tweets about the other coins, suggesting a relatively lower level of interest or engagement in discussions related to it. For Shiba, our observations indicate a prevalent sense of hope and an increased use of filler words compared to the other cryptocurrencies. This heightened expression of hope suggests a more optimistic sentiment surrounding Shiba when contrasted with the other coins. Additionally, the frequent use of filler words, including expressions like wow, sooo, and youknow signifies a more conversational and engaged discourse. This linguistic pattern may reflect a greater level of enthusiasm and interaction among Shiba enthusiasts.
This analysis includes words commonly used in social media and text messaging, such as bae, lol and basic punctuation-based emoticons like 🙂 and ;)65,71. This mode of communication is widely employed by netizens during computer-mediated communication (CMC). In the context of cryptocurrency discussions, which predominantly transpire on online forums, social media platforms, and chat groups, it is customary for participants to incorporate netspeak into their interactions. Through the analysis of netspeak, researchers can understand more the degree of user engagement and interaction. Notably, the adoption of terms such as HODL (a deliberate misspelling of hold, indicating a long-term investment strategy) or moon (indicating an expectation of significant price increases) serves as meaningful pointers to user sentiment and active participation in discussions. In the obtained results, Matic stands out prominently with a notably high netspeak score, signaling the prevalence of internet-specific expressions and informal language related to it. The results can be found in Table3. Fillers (wow, sooo, youknow) are non-essential sounds, words, or phrases, such as well, erm or hmm commonly used in speech to occupy pauses and maintain the flow of conversation without altering its meaning65,72,73. The filler analysis results highlight that Shiba and Dogecoin exhibit higher scores in this category compared to the other cryptocurrencies, with scores ranging between 0.02 and 0.04 for the remaining coins, as depicted in Table3. In the sentiment analysis, its clear that Fantom distinguishes itself with a notably elevated positive score in comparison to the other cryptocurrencies. A consistently positive sentiment can enhance investor confidence, attract new stakeholders, and contribute to a more favorable market perception. Table3 presents the remaining outcomes for the other cryptocurrencies.
Table4 provides a detailed sentiment analysis, encompassing positive, neutral, and negative percentages for various digital coins. In the world of cryptocurrency investments, its common for investors to assess public sentiment before making their decisions, as highlighted in prior research30. Consequently, sentiment analysis has gained substantial importance on cryptocurrency markets74. Studies have shown that tweets expressing positive emotions wield substantial influence over cryptocurrency demand, while negative sentiments can have the opposite effect32,33.
Analyzing the data in Table4, it becomes apparent that Fantom distinguishes itself by displaying a notably higher positive sentiment percentage in comparison to its digital counterparts, which strongly suggests an elevated degree of interest and enthusiasm among investors towards this digital coin.
Examining opinions involves another aspect known as emotion detection. In contrast to sentiment, which can be positive, negative, or neutral, emotions offer richer categorization over personality traits by revealing experiences of joy, anger, and more. Automated methods for emotion detection have been developed to enhance the analysis of individual sentiments. The primary goal of emotion analysis is to identify the specific words or sentences conveying emotions75. To achieve such analysis, we employed the NRCLex library to extract and categorize emotions from text24. NRCLex is a Python library designed for natural language processing and sentiment analysis. The acronym stands for Natural Resources Canada Lexicon, and it is particularly focused on assessing sentiment in text based on word associations. NRCLex is built upon a lexicon that assigns sentiment scores to words, allowing users to analyze the sentiment of individual words, sentences, or entire documents76. Table5 provides the outcomes of our emotion analysis, revealing a narrow range of results for various emotions: Anger (0.02-0.04), Surprise (0.01-0.02), Sadness (0.01-0.03), Disgust (0.01-0.02), and Joy (0.02-0.04). These consistent findings suggest that most of the coins evoke similar emotional responses, highlighting their emotional proximity.
In contrast, when it comes to emotions such as Fear and Trust, there are more noticeable differences between the coins. For instance, when examining the sentiment of Cardano, the fear score is 0.0324, while the trust score is higher at 0.1252. Similarly, for Ripple, the fear score is 0.0416, with a trust score of 0.1172. The scores provide a difference in the emotional tones associated with these cryptocurrencies, indicating the levels of fear and trust expressed in the analyzed content.
Furthermore, the emotion of Anticipation stands out with higher scores in tweets, indicating that many users are keen on anticipating the future of these coins. Notably, Dogecoin (0.3752) and Shiba (0.3467) generate more anticipation among users when compared to the other coins.
In this section, we pay attention to the readability of data, utilizing metrics such as the Flesch Reading Ease25, Flesch-Kincaid Grade Level26, Gunning Fog Index27, and Dale-Chall Readability Score28. Assessing readability helps distinguish between text that is straightforward to grasp and text that is complex and demands a high level of education or intelligence to comprehend. Numerous readability metrics exist for text evaluation, and we have chosen to employ the above four measures as the most widely recognized tests to assess tweets.
Table6 presents the significant differences in readability scores across tweets related to nine different digital coins.
The Flesch Reading Ease score provides an indication of how easily a text can be understood, with higher scores indicating greater readability. Flesch Reading Ease score can be observed in Fig.5. The Flesch-Kincaid Grade Level is a metric that estimates the educational grade level required to understand a piece of text based on factors like sentence length and word complexity. Analyzing the readability scores for the tweets related to each digital coin shows the linguistic complexity employed in discussions surrounding these coins. The presence of significant differences in readability scores suggests variations in the accessibility and comprehension levels required to engage with these tweets. Negative scores in some readability metrics, such as the Flesch Reading Ease and Flesch-Kincaid Grade Level, indicate higher levels of complexity, while positive scores indicate greater ease of comprehension. Refer to Fig.6 for the necessary details to assess the readability levels of the specified analyses (Flesch-Kincaid Grade Level, Gunning Fog Index, Dale-Chall Readability Score). Table6 provides evidence on the fact that Dogecoin possesses a notably lower score in Flesch Reading Ease compared to the other cryptocurrencies, which suggests that the communication pertaining to Dogecoin might present hurdles in accessibility and comprehension for the typical reader. Getting rid of such readability obstacles have the potential to amplify the effectiveness of communication, expand audience involvement, and cultivate heightened comprehension and acceptance of cryptocurrencies among varied stakeholders. This observation aligns with Fig. 577, where we notice a pronounced level of complexity in comprehending tweets related to Dogecoin. To gain a better understanding of the varied readability levels, its essential to consider both Fig.578,79 and Table6. When examining the Flesch-Kincaid Grade Level and Dale-Chall Readability in Table6, Dogecoin emerges with higher values compared to the other cryptocurrencies, signifying an average grade level and a college reading level, respectively. Furthermore, an examination of the results pertaining to the Gunning Fog Index, as depicted in Table6 and Fig.6, reveals that Ethereum stands out with a higher score. This observation implies that understanding tweets related to Ethereum requires a reading comprehension level equivalent to a college education.
Flesch reading ease score.
Dale-Chall Readability Score, Gunning Fog Index, Flesch-Kincaid Grade Level.
Follow this link:
Psycholinguistic and emotion analysis of cryptocurrency discourse on X platform | Scientific Reports - Nature.com