Journal of Science &Technology Metrics Sandra M K, Rupesh Kumar A 7 3 2026 https://doi.org/10.6025/jstm/2026/7/3/150-160 https://www.dline.info/jstm/fulltext/v7n3/jstmv7n3_3.pdf Purpose: This study examines the contextual and sectional relevance of citations received by scholarly publications in computational linguistics using Scite. It further investigates the relationships between supporting, contrasting and mentioning citations and the corresponding Scopus citation counts. Methodology: Both conventional and contextual citation-analysis techniques were employed. Scientific publications on computational linguistics were retrieved from the Scopus database. Contextual citation data were subsequently obtained from the smart citation index Scite.ai, an emerging classified citation database built with machine learning, artificial intelligence and natural language processing. A total of 15,607 documents containing classified citation statements were included. Data were analysed in Microsoft Excel and statistical tests were performed with SPSS. Results: The analysis reveals a steady rise in both publications and citations. The 15,607 scholarly works collectively received 3,872 supporting, 375 contrasting and 343,855 mentioning citations corresponding to 1.09 % agreement, 0.11 % disagreement and 97.23 % neutral research engagement. Mentioning citations dominate every year, reflecting a tendency among scholars to cite neutrally without explicit endorsement or critique. Although contrasting citations are far fewer than supporting or mentioning citations, they remain significant because they signal critical scrutiny and inconsistencies in the published literature. Among the ten journals with the highest publication counts, Transactions of the Association for Computational Linguistics leads in both supporting and contrasting citations. Spearman’s rank-correlation tests show a strong positive association between mentioning citations and Scopus citations ( = .839, p = .000), a weak positive association with supporting citations (  = .336,  = .000) and a still weaker positive association with contrasting citations ( = .153, p = .000). Sectional citation analysis indicates that the largest number of citations (6,031) fall into the “Other” category, underscoring the diversity of structural conventions in scientific communication. Conclusion: The investigation demonstrates the value of contextual citation analysis applied to computational linguistics research. The approach can be extended to author-, journal- and institutionallevel assessments to reveal genuine research impact. Contextual citation analysis offers a novel perspective that moves beyond the traditional assumption that “all that glitters is gold.”