Scholarlink

The integration of artificial intelligence (AI) models into academic publication management systems presents a promising solution to the difficulties of structuring composite data, manually input by authors and records retrieved from application programing interfaces (API). Current systems face barriers in accurately matching records from diverse scientific databases, such as Scopus or Web of Science, leading to inefficiencies and inaccuracies. In this paper we discuss existing bibliometric solutions at the University of Economics – Varna, Bulgaria and tackle different algorithmic approaches for integrating automated API communication. The research explores the use of AI, specifically the DeepSeek-R1 model based on Qwen or Llama fine-tuning distillations, to enhance automated validation and discoverability of new research for the purpose of institutional accreditation reports. We address two primary challenges: identifying duplicate entries with heterogenous structures and improving bibliometric data management. The proposed solution involves the calculation of similarity scores using metrics like locality-sensitive hashing. The development of AI-powered tools and their integration with academic systems offers a novel approach to overcoming current limitations. This study highlights the potential benefits of using DeepSeek-R1 for enhancing both the accuracy and scalability of publication and citation data management.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Academic Profile Management: Benchmarking DeepSeek-R1 for Publication and Citation Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Boris Bankov,
  • Silvia Parusheva,
  • Olga Marinova,
  • Petya Strashimirova,
  • Denitsa Petkova

摘要

The integration of artificial intelligence (AI) models into academic publication management systems presents a promising solution to the difficulties of structuring composite data, manually input by authors and records retrieved from application programing interfaces (API). Current systems face barriers in accurately matching records from diverse scientific databases, such as Scopus or Web of Science, leading to inefficiencies and inaccuracies. In this paper we discuss existing bibliometric solutions at the University of Economics – Varna, Bulgaria and tackle different algorithmic approaches for integrating automated API communication. The research explores the use of AI, specifically the DeepSeek-R1 model based on Qwen or Llama fine-tuning distillations, to enhance automated validation and discoverability of new research for the purpose of institutional accreditation reports. We address two primary challenges: identifying duplicate entries with heterogenous structures and improving bibliometric data management. The proposed solution involves the calculation of similarity scores using metrics like locality-sensitive hashing. The development of AI-powered tools and their integration with academic systems offers a novel approach to overcoming current limitations. This study highlights the potential benefits of using DeepSeek-R1 for enhancing both the accuracy and scalability of publication and citation data management.