EULingDiv

Altough the European Union is commited to linguistic diversity by recognizing 24 offical EU-languages and actively promiting multilingualism, there have not been much effort vested into quantitatively asessing the linguistic diversity of the union. However, two eurobarometer surveys have been conducted that tangentially surves this purpose: one in 2012, and one in 2024. In the surveys, question are posed pertaining to the native language, and first to third other language, which we interpret as corresponding to L1 to L4, thus potentially allowing for more accruate models of linguistic diversity that account for multilingualism. To allow for easier anlysis of the data, we have merged and structured the data pertaining to spoken language across the surveys into the dataframe EU_country_speakers_2012_2024.csv. The dataframe is structured in a long format with country-language-year-group-speakers pentuplets. We have not converted the numbers into proportions and derived any formal measures, since there are many analysis-specific aspects necessary to account for, e.g. the weighting given to L1 contra L2, L3, etc. Furthermore, the survey is sample based, and by summing the L1-speakers of each country, one arrives at the sample size. As such, consideration needs to be made regarding the uncertainty of potential diversity indices derived from the data. While we have added ISO-codes to make the dataset more interoperable, one should be wary of the presence of macrolanguages. For example, the survey designates speakers of "Arabic" and "Albanian", but does not specify the varieties such as Tosk or Gheg Albanian. We have kept with this and added the macroidentifyer of "ara" and "sqi" in such cases. Furthermore, the languages included in the survey of 2012 and 2024 does not necessarily fully overlap due to how the surveys were designed per design of the survey constructors. The dataset contains the following columns: country_code: 2-letter ISO-3166 code denoting the country that was surveyed. (character) country_name: Commonly used country names corresponding to the country code - does not follow any particular standards. (character) ISO6393: 3-letter ISO639-3 code denoting the language that was surveyed. (character) Language name: Language name used by the original surveys. (character) Speaker_type: Categorical variable denoting if the language is the first (L1), second (L2), third (L3) or fourth (L4) language of the speakers. (character) number_of_speakers: Variable denoting the number of speaker of a variety in a given country according to a given type. (double) year: Denotes which year and survey the data is sourced from (integer) The code developed to construct the dataset can be found at: https://github.com/Eszettfors/EULingDiv ------ v.1.1 fixed an error regarding the L4 speaker numbers in 2024.尽管欧盟通过认可24种欧盟官方语言并积极推动多语言使用，致力于保障语言多样性，但针对欧盟语言多样性开展量化评估的投入却相对不足。不过已有两项欧洲晴雨表（Eurobarometer）调查间接围绕这一目标展开，分别于2012年与2024年实施。调查中设置了关于母语以及第一至第三外语的问题，我们将其对应为L1至L4，这为构建考虑多语言因素的更精准语言多样性模型提供了可能。 为便于数据分析，我们将两次调查中涉及口语使用的相关数据进行整合与结构化处理，得到数据框（dataframe）文件EU_country_speakers_2012_2024.csv。该数据框采用长格式结构，包含国家-语言-年份-群体-说话者数量五元组。由于存在诸多需根据分析场景考量的特定因素（例如为L1、L2、L3等语言赋予不同权重），我们未将数值转换为占比或推导任何正式的多样性测度指标。此外，本次调查基于抽样设计，通过累加各国的L1使用者人数即可得到样本量，因此在基于该数据推导潜在多样性指数时，需充分考虑其不确定性。为提升数据集的互操作性，我们添加了ISO代码，但需注意其中存在宏语言（macrolanguage）的情况。例如，调查将使用者划分为“阿拉伯语”和“阿尔巴尼亚语”群体，但未指明其方言变体（如托斯克阿尔巴尼亚语与盖格阿尔巴尼亚语），对此我们保留了原分类，并为这类情况添加了宏语言代码“ara”与“sqi”。另外，由于两次调查的设计方案不同，2012年与2024年调查所涵盖的语言未必完全重合。 本数据集包含以下字段： country_code：表示受访国家的2位字母ISO-3166代码（字符型） country_name：与国家代码对应的常用国名，未遵循特定命名标准（字符型） ISO6393：表示受访语言的3位字母ISO 639-3代码（字符型） Language name：原始调查中使用的语言名称（字符型） Speaker_type：分类变量，用于标识该语言为使用者的第一（L1）、第二（L2）、第三（L3）或第四（L4）语言（字符型） number_of_speakers：表示特定类型使用者在对应国家中使用某语言变体的人数（双精度浮点型） year：数据来源的调查年份（整数型） 本数据集的构建代码可访问以下链接获取：https://github.com/Eszettfors/EULingDiv --- 版本v1.1：修复了2024年调查中L4使用者人数相关的错误。