Building a test collection for complex document information processing

By David D. Lewis, Gady Agam, Shlomo Engelson Argamon, Ophir Frieder, D. Grossman, J. Heard. 2006

326 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Evaluation & Benchmarks, Information Extraction

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Furu Wei
2. Chenliang Li
3. Shaohan Huang
4. Ming Yan
5. Bin Bi
6. Songfang Huang
7. Wei Wang
8. Ming Zhou

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Furu Wei
3. Dacheng Tao
4. Andrew Zisserman
5. Cordelia Schmid
6. Andrea Vedaldi
7. Jitendra Malik
8. Trevor Darrell

In Evaluation & Benchmarks

1. Douglas W. Oard
2. Yupan Huang
3. Fei Huang
4. Chenliang Li
5. Lei Cui
6. Raghavan Manmatha
7. Dinei A. F. Florêncio
8. Ming Yan

In Information Extraction

1. Fei Huang
2. Wanxiang Che
3. Songfang Huang
4. Furu Wei

Citations in the corpus

• LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking (2022, background): “0 is a large-scale scanned document image dataset, which contains about 11 million document images and can split into 42 million pages [24].”
• LayoutLMv2: Multi-modal Pre-training for Visually-rich Document Understanding (2020, methodology): “Following LayoutLM, we use IIT-CDIP Test Collection (Lewis et al., 2006) as the pre-training dataset.”
• OCR-Free Document Understanding Transformer (2021, methodology): “We use IIT-CDIP [32], which is a set of 11M scanned english document images.”
• LayoutLM: Pre-training of Text and Layout for Document Image Understanding (2019, methodology): “ecause they can be considered as a special case where OCR is not required, thus they are out of the scope of this paper. Specifically, the LayoutLM is pre-trained on the IIT-CDIP Test Collection 1.02 [14], which contains more than 6 million scanned documents with 11 million scanned document images. ”
• Evaluation of deep convolutional nets for document image classification and retrieval (2015, methodology): “The performance of the proposed approach was evaluated on two versions of the IIT-CDIP test collection [17].”
• TREC 2006 Legal Track Overview (2006, methodology): “The IIT CDIP project subsequently reformatted the metadata and OCR, combined the metadata with a slightly different version obtained from UCSF in July 2005, and discarded some documents with formatting problems, to produce the IIT CDIP 1.0 collection (Lewis, et. al 2006).”
• DocFormer: End-to-End Transformer for Document Understanding (2021, methodology): “In order to pre-train DocFormer we use a subset of 5 million pages from the IIT-CDIP document collection [31] for pre-training.”
• StructuralLM: Structural Pre-training for Form Understanding (2021, methodology): “Following LayoutLM, we pre-train StructuralLM on the IIT-CDIP Test Collection 1.0 (Lewis et al., 2006).”