A Decade of News Forum Interactions: Threaded Conversations, Signed Votes, and Topical Tags

Area of Science:

• Computational Social Science
• Digital Humanities
• Natural Language Processing

Background:

• Online news platforms generate vast amounts of user-generated content.
• Analyzing online discourse requires large-scale, longitudinal datasets with rich metadata.
• Existing datasets may lack sufficient detail or long-term scope for comprehensive analysis.

Purpose of the Study:

• To release a large-scale, longitudinal dataset of user activity from DerStandard, an Austrian newspaper.
• To enable research on online discussion dynamics, network structures, and semantic analysis.
• To provide a privacy-preserving resource for computational social science research.

Main Methods:

• Collected and structured ten years (2013-2022) of user comments and votes.
• Anonymized user identifiers using salted hash functions.
• Generated pre-computed vector representations of comment texts using a state-of-the-art embedding model.

Main Results:

• A dataset comprising over 75 million user comments and 400 million votes.
• Detailed metadata on articles and user interactions, including structured conversation threads.
• Privacy-preserving release of comment data via vector representations, excluding raw text.

Conclusions:

• The DerStandard dataset is a valuable, reusable resource for studying online discourse in German.
• The dataset facilitates novel research in computational social science and digital humanities.
• The methodology ensures user privacy while enabling rich data analysis.