940,000 Online Circulations of 26 Iconic Photographs with Text Embeddings

Utrecht University researcher Thomas Smits compiled this dataset for the article 'Distant reading patterns of iconicity in 940.000 online circulations of 26 iconic photographs'. The material includes tabular metadata on webpages hosting reproductions of 26 iconic photographs, identified via the Google Cloud Vision API, and associated Doc2Vec text embeddings. The dataset supports analysis of how iconic images circulate and are discussed online.

Use Cases

• Clustering webpage text descriptions based on the provided Doc2Vec embeddings.
• Analyzing the diachronic prominence of image-related discourse using the included cluster plots.
• Studying the association between computer vision labels and webpage language metadata.
• Investigating the network of online reproductions for iconic photographs based on URL metadata.

Strengths

• Focuses on 26 pre-identified iconic photographs, providing a curated subject set.
• Includes approximately 940,000 online circulations, indicating a large-scale collection.
• Contains multiple data types: tabular URL metadata, computer vision labels, and text embeddings for analysis.
• Data cleaning involved manual relevance checks and a computer vision algorithm (SIFT) to reduce noise.

Limitations

• Row count and file size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The date parser failed for some webpages, resulting in incomplete temporal metadata.
• Last update date is unknown; freshness unverified.

Provenance

Source

Utrecht University

Collection Method

Webpages were identified using the Google Cloud Vision API to find reproductions of 26 iconic photographs. Metadata was scraped and text from pages was processed into Doc2Vec embeddings.