Get started
Work

Who Writes Persian NLP? An Inward-Facing Research Community in International Perspective, 2005–2025

Year: 2026
Type: preprint
Abstract: Purpose. Persian is spoken by tens of millions of people, yet the research community that produces natural language processing (NLP) work on it has not been measured. We ask how fast it has grown, how internationally connected it is, how visible its output is, who writes it, and where it appears, in comparison with eight other languages. Design and methods. From OpenAlex we retrieved English-language articles, preprints and conference papers classified in the Artificial Intelligence subfield, 2005–2025, whose titles name one of nine languages (Persian, Arabic, Turkish, Urdu, Hebrew, Hindi, Indonesian, Vietnamese, Thai) and whose titles or abstracts contain NLP-related words (663 Persian works). We compared growth (Poisson-based intervals and a test against the pooled other languages), home-country and international co-authorship, and the share of works in the citation top-10% (Wilson intervals, Fisher exact tests), repeated each comparison in a pre-break window (2015–2021) to guard against OpenAlex's post-2021 coverage break, and analysed 614 Persian records by author-country group, institution and venue. Findings. Persian output grew by a factor of 1.53 (95% CI 1.27–1.84) between 2016–20 and 2021–25 against 1.96 for the other eight languages pooled (p = 0.011), and its share of the nine languages' output fell from 10.0% in 2010 to 5.5% in 2024; the pre-break comparison points the same way but is not conclusive (1.43 against 1.78, p = 0.067). 69.2% of Persian works have an Iran-affiliated author, and 10.2% are internationally co-authored, against 16.4% for the pooled other languages (p < 0.001; 8.9% against 14.0% before the break, p = 0.020). 6.7% of Persian works reached the citation top-10% against 16.5% (p < 0.001), but this gap is shared with Hebrew, Indonesian and Vietnamese and does not distinguish Persian from them; Arabic and Urdu, which span several countries, are the outliers. Of 614 records, 59.1% have Iranian authors only, 8.0% mix Iranian and foreign authors, 9.6% have only foreign-affiliated authors and 23.3% lack affiliation country data; works by foreign-affiliated authors only are more likely to be above world-average impact (odds ratio 4.58, 95% CI 1.93–10.86; n = 25). Ten of the twelve leading institutions are in Tehran, and the corpus is carried mostly by journals and repositories. Originality. To our knowledge, this is the first quantitative portrait of the community behind Persian NLP, benchmarked against comparable language communities, and it separates real differences from those that single-country communities share, while documenting two data-quality hazards (missing affiliation country and the 2022 coverage break) that change naive readings.
DOI: 10.5281/zenodo.23448579
Language: English
Open Access status: green
OpenAlex ID: W7222882250
Cites:
Cited by:
Related to: