This dataset contain 825 tweet instances of Indonesian-English, corresponding to four NLP tasks, i.e., tokenization, language identification, lexical normalization, and word translation. Data for lexical normalization task is curated in MultiLexNorm (already in Nusa Catalogue), but other tasks are not. Tokenization for social media data is not as trivial as splitting the token using white space delimiter. In this data, language identification is performed in token-level granularity.
NusaCatalogue: https://indonlp.github.io/nusa-catalogue/card.html?id_en_code_mixed