The frGLOBE corpus is part of the BFSU Global Corpus Cluster project, aimed at creating a million word balanced corpus of French written language. This project refers to the sampling method of the Brown corpus to construct the GLOBE series corpus for bilingual or multilingual comparative research. The frGLOBE corpus focuses on collecting original French texts first published or published between 2013 and 2023, with a design scale of 1 million words.

1 History of French Corpus Construction

This article reviews the development of the French language corpus, starting from the compilation of the "French Encyclopedia" in the 1960s, and introduces the characteristics and functions of its digital version. Subsequently, the chapter provides a detailed description of the establishment and content of the French language database, including the types of texts it includes and its search function. In addition, other types of French corpora were also mentioned, such as large-scale reference corpora, oral corpora, diachronic corpora, learner corpora, and specialized corpora. The chapter points out that although the French corpus has abundant resources, there are still shortcomings in terms of balanced construction, such as excessive concentration of literary texts and lack of comparative research with other language corpora such as English and Chinese. The frGLOBE corpus project aims to address these issues and provide a supplement to the balance and diversity of the French corpus.

2. Creation of frGLOBE Corpus

The frGLOBE corpus is a million word balanced corpus that uses the Brown corpus model to collect written French language versions after 2010. This corpus includes three versions: raw corpus, part of speech encoding, and word form restoration, annotated using the spaCy tool. The corpus has been uploaded to the CQPweb platform of Beijing Foreign Studies University, providing various analysis functions and accompanied by a French word list and phrase list. The sampling scheme refers to the Brown corpus, covering four genres: news, general, academic, and novel, subdivided into 15 subcategories, and adjusted appropriately according to French culture. Corpus collection includes text collection, input, and meta information annotation. The authors are all native French speakers, and the text is original. The corpus contains 500 texts of approximately 2000 words, stored as 1034 sub files, following a specific naming format. The metadata is recorded in an Excel spreadsheet. All three versions of the corpus can be used on the CQPweb platform, supporting simple queries and CQP syntax retrieval.

3 Research Prospects

The frGLOBE corpus provides rich resources for French language research, supporting the study of vocabulary, phrases, syntax, and discourse features, and can be used for teaching and dictionary compilation. The corpus includes four types of texts: news, general, academic, and novel, which are suitable for stylistics and domain variation research. The correspondence with the Brown family corpus and other GLOBE corpora facilitates comparative research, such as French English, French Chinese, and typological studies. The corpus can be combined with the French language database to promote research on the historical changes of French language. Subsequent versions will add updated corpus metadata, reduce the proportion of texts before 2014, consider including social media texts, and more accurately reflect the use of French written language in the past decade. If conditions permit, oral language materials can be added to conduct comparative research on oral and written language. The frGLOBE corpus can be used together with other GLOBE corpora to construct multilingual parallel comparable corpora, promoting comparative analysis and research in multiple languages.

4 Conclusion

The frGLOBE corpus provides a balanced corpus for contemporary French written language, filling the gap in French written language resources since 2014. This corpus adopts the Brown corpus sampling scheme, supporting independent research and combined use with other corpora. By providing online search and vocabulary functions through the "BFSU CQPweb Multilingual Corpus Platform", we encourage co construction and sharing, and promote the development of French teaching and research.