Analysis of the DBLP Publication Classification Using Concept Lattices

Saleh Alwahaishi, Jan Martinovic, Václav Snásel, Milos KudelkaDATESO 2011

TL;DR

This paper introduces a novel approach to classifying scientific journals using Formal Concept Analysis (FCA) on the DBLP dataset. By analyzing journal aims and scopes, FCA reveals hidden relationships between journals and their topics, offering a more intuitive and structured way to understand the landscape of scientific publications.

What Problem Does This Paper Address?

Classifying scientific journals is crucial for researchers to navigate the vast academic landscape, discover relevant venues, and understand the interconnections between different fields. However, traditional classification methods can be rigid and fail to capture the nuanced relationships and overlapping scopes of journals. The DBLP dataset, a comprehensive bibliography of computer science publications, presents an opportunity to analyze these relationships, but its raw data lacks explicit hierarchical classifications. This paper aims to bridge this gap by using FCA to create a visual and structured classification of journals based on their stated aims and scopes.

Methods

The core methodology revolves around Formal Concept Analysis (FCA). The process begins with pre-processing the DBLP dataset to extract the "aim and scope" information for each journal. This information is then used to identify key topics (attributes) covered by each journal. A formal context is constructed, representing journals as objects and topics as attributes, with a binary value indicating whether a journal covers a specific topic. FCA is then applied to this formal context to generate a concept lattice. This lattice visually represents concepts (combinations of journals and topics) and their hierarchical relationships, revealing how journals and topics are related. To manage the scale of the dataset, journals with an impact factor of 3.0 or above were selected for the case study, resulting in a reduced set of 18 journals.

Datasets

The primary dataset used is the DBLP (Digital Bibliography & Library Project) publication database. For the specific analysis, a subset of journals from DBLP with an impact factor of 3.0 or higher was selected, totaling 18 journals. The "aim and scope" information for these journals was manually extracted from their respective websites.

Key Insights and Ideas

  • FCA as a Visualization Tool for Journal Classification: FCA provides a powerful way to visualize the complex relationships between journals and their thematic coverage, moving beyond simple keyword matching.
  • Revealing Implicit Structures: The concept lattice uncovers implicit structures and contextual information about how journals are associated, based on their stated aims and scopes.
  • Topic Granularity and Relationships: The analysis demonstrates how FCA can identify sub-concepts and hierarchical relationships between topics. For instance, "DF" (Data Format) being a sub-concept of "AIIP" (Artificial Intelligence and Image Processing) suggests a more specific relationship within the broader field.
  • Complementary to Traditional Ontologies: Concept lattices offer a flexible way to represent less rigorous associations and intuitive links, complementing traditional, more rigid ontologies.

Key Results

The paper successfully constructs a concept lattice for a selected set of 18 high-impact journals from the DBLP dataset. This lattice visually represents the relationships between journals and their covered topics. For example, the paper highlights how the topics "CTM" (Computer Theory and Mathematics) and "DIP" (Data and Information Processing) are shown as related because they share a similar classification context. The lattice also allows for the identification of specific relationships, such as "DF" being a more specific topic than "AIIP." The authors suggest that this approach helps researchers classify journals more efficiently and can support the emergence of new knowledge by making concept relations explicit.

Limitations and Future Work

  • Manual Extraction of Aims and Scopes: The extraction of journal aims and scopes was done manually, which can be time-consuming and potentially introduce human bias.
  • Limited Dataset Size for Case Study: The case study focused on a subset of 18 journals with high impact factors. A broader analysis across the entire DBLP dataset would provide a more comprehensive view.
  • Future Work: The authors suggest that this approach has the potential to support the emergence of new knowledge by identifying concept relations, making them explicit, and enabling researchers to inspect these relationships. Further exploration of combining FCA with other representations like description logics or conceptual graphs is also hinted at.

Summary updated

BibTeX
@inproceedings{alwahaishi2011analysis,
  author = {Saleh Alwahaishi and Jan Martinovic and Václav Snásel and Milos Kudelka},
  title = {{Analysis of the DBLP Publication Classification Using Concept Lattices}},
  booktitle = {DATESO},
  year = {2011},
}
Figure