Using ColPali embeddings of 3,600 textbook page images, cosine similarity beats dot product, Euclidean, and Manhattan distances on top-5 retrieval, but only reaches 0.51 precision@5 without a text-only baseline.
Introduction to and hands-on use cases with hathitrust research center’s extracted features 2.0 dataset
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
dataset 1
citation-polarity summary
fields
cs.IR 1years
2025 1verdicts
CONDITIONAL 1roles
dataset 1polarities
background 1representative citing papers
citing papers explorer
-
Vector embedding of multi-modal texts: a tool for discovery?
Using ColPali embeddings of 3,600 textbook page images, cosine similarity beats dot product, Euclidean, and Manhattan distances on top-5 retrieval, but only reaches 0.51 precision@5 without a text-only baseline.