Abstract
Twitter has become the de facto information sharing and communication platform. Given the factors that influence language on Twitter—size limitation as well as communication and content-sharing mechanisms—there is a continuing debate about the position of Twitter's language in the spectrum of language on various established mediums. These include SMS and chat on the one hand (size limitations) and email (communication), blogs and newspapers (content sharing) on the other. To provide a way of determining this, we propose a computational framework that offers insights into the linguistic style of all these mediums. Our framework consists of two parts: a set of linguistic features to quantify the language of a given medium, and a flexible factorization framework, soclin, which conducts a psycholinguistic analysis with the help of an external cognitive and affective knowledge base. Applying this framework to various corpora, we present several key insights: (1) Twitter's language is surprisingly more conservative and less informal than SMS and online chat; (2) Twitter users appear to be developing linguistically unique styles; (3) Twitter's usage of temporal references is similar to SMS and chat; and (4) Twitter has less variation of affect than other more formal mediums. The language of Twitter can thus be seen as a projection of a more formal register into a size-restricted space.
Introduction
Since its inception, linguists and the public have debated where Twitter sits on the "linguistic spectrum." Is it a digital version of spoken conversation, like SMS and online chat, or a compressed form of professional writing, like blogs and newspapers? The challenge in answering this lies in the nature of "micro-text." Because tweets are so short, traditional counting-based linguistic tools often fail due to data sparsity—there simply aren't enough words in a single tweet to reliably categorize its emotional or cognitive state. This study seeks to bridge that gap by providing a principled, quantitative framework to place Twitter within the broader context of human communication.
Methodology
The research utilizes a two-part computational framework to analyze seven large-scale datasets, including 45 million tweets from the 2012 Firehose, the NUS English SMS corpus, NPS chatrooms, Enron emails, ICWSM blogs, Slate magazine articles, and Reuters news data.
- Linguistic Feature Analysis: Standard metrics were measured, including Word Frequency (readability), Lexical Density (the ratio of information-carrying words to total words), and grammatical markers like Personal Pronouns and Intensifiers.
- SOCLIN (Social Linguistic Factorization): This matrix factorization framework decomposes a term-document matrix into latent "psycholinguistic dimensions." To solve the sparsity problem of short tweets, it uses the LIWC (Linguistic Inquiry Word Count) dictionary as a "prior" to supervise the learning process, allowing the model to infer the "affect" (emotion) or "cognition" of a tweet based on the context of similar documents.
The "Formal Projection" Hypothesis
The study concludes that Twitter is not merely "SMS 2.0." Instead, it is a projection of formal registers (like news and blogs) into a size-restricted space. Users are not simply being lazy; they are adapting professional-style communication to fit a 140-character limit.
Style Over Convenience
One of the most striking findings is that users do not automatically swap "you" for "u" or "I" for "i" simply to save space. Instead, the data suggests that linguistic choice on Twitter is a matter of personal branding. Users tend to pick a specific style—either formal or informal—and maintain it consistently across their posts.
Information vs. Interaction
Twitter's linguistic profile aligns more closely with information-sharing mediums than interpersonal ones. For instance, Twitter uses significantly more third-person pronouns than SMS or Chat. This indicates that the platform is primarily used for broadcasting news or status updates rather than the one-to-one back-and-forth interaction typical of private messaging.
Key Results
- Lexical Density: Twitter’s lexical density (0.47) is much closer to Slate magazine (0.48) than to SMS (0.42) or Chat (0.40), meaning tweets are packed with more information-carrying words like nouns and verbs.
- Vocabulary Sophistication: Twitter uses more "sophisticated" (less frequent) words than SMS and Chat, aligning more closely with the vocabulary found in emails and blogs.
- The Positivity Bias: Twitter exhibits higher "positive affect" and lower variation in emotion compared to other mediums. As a public broadcast medium, users appear to curate their personas to appear more positive than they might in private communication.
- Cognitive Certainty: Twitter contains more words related to "certainty" (e.g., always, never) and fewer "tentative" words (e.g., maybe, guess) than SMS, suggesting it is a platform for strong opinions rather than exploratory discourse.
Conclusion and Future Work
While the study is limited by the temporal differences between datasets (2006–2012) and the specific demographic skew of early Twitter users, it provides a robust quantitative argument for Twitter's relative formality. Future research will focus on diachronic analysis to see how Twitter’s language has evolved over time and how specific social network structures influence the linguistic formality of a user's feed.
Summary updated