Exploration of a citation network based on agreement type

This project aims to enable alternative exploration of a corpus of scientific publications—regardless of its size—based on the nature of the relationships between the references (agreement, disagreement, or neutral mention). The goal is to produce visualizations or exploratory methods that challenge the quantitative significance of citation counts.

Issues

Challenging the primacy of the absolute number of citations in evaluating research outcomes on a search engine requires the definition of qualitative criteria. Based on the observation that citation discordance is more pronounced in the social sciences and humanities (Lamers et al. 2021), this project aims to develop an analytical tool capable of characterizing the nature of citation relationships between scientific publications.

Technical challenges

• Manual annotation of a large number of articles with citation types to evaluate the alignment between the bibliography and in-text citations, and to classify citation types.
• Ability to refine a model if the pre-trained classifiers prove insufficient.

Research activities

• Development of a Python program for data extraction and visualization using HTML/JS.
• Evaluation of data extraction and classification using a manually created dataset.
• Scaling up the program for deployment with partners Érudit and Isidore.

Deliverables

• A classifier that aligns the bibliography with inline citations and identifies the citation type.
• Visualizations across multiple delimited corpora.