Kazakhstan focused scientific publications from Semantic Scholar

This dataset contains 10,000 Kazakhstan-related English-language scientific articles from 2012–2025 with metadata (paperId, title, abstract, year, DOI) and a derived set of 14,460 abstract pairs for duplicate detection research. Pairs were formed using TF-IDF and cosine similarity to select real near-duplicates and non-duplicates, plus synthetic paraphrases generated by controlled word replacement, resulting in a balanced mix of positive (duplicate) and negative (non-duplicate) pairs.