Language Statistics at Different Spatial, Temporal, and Grammatical Scales

Area of Science:

• Computational Linguistics
• Quantitative Linguistics
• Sociolinguistics

Background:

• The availability of large datasets has propelled statistical linguistics.
• Twitter data offers a rich resource for analyzing real-time language use.

Purpose of the Study:

• To investigate rank diversity in English and Spanish using Twitter data.
• To examine the influence of temporal, spatial, and grammatical scales on language variation.
• To quantify universal language statistics and identify sources of variation.

Main Methods:

• Analysis of Twitter data from 2014 across eight countries.
• Investigation of word ngrams (1-grams to 5-grams) across temporal (3-96h) and spatial (3km-3000km) scales.
• Examination of rank diversity curves and statistical properties of Twitter-specific tokens (emojis, hashtags, mentions).

Main Results:

• Rank diversity shows similarity at the 1-gram level across languages, countries, and scales.
• Increased grammatical complexity (higher ngrams) leads to more pronounced variations influenced by temporal, spatial, linguistic, and national factors.
• Twitter-specific tokens exhibit a sigmoid pattern in their rank diversity function.

Conclusions:

• Grammatical scale is a key driver of language rank diversity variation.
• While universal patterns exist, language use on Twitter is shaped by context (time, location, nationality).
• The study quantifies language statistics and highlights factors influencing linguistic diversity in digital communication.