Unlocking Maps: Automatic and Streamlined Metadata Creation for Digital Collections

FAIN: PR-253386-17

University of Southern California (Los Angeles, CA 90089-0012)
Deborah Ann Holmes-Wong (Project Director: June 2016 to June 2019)

The evaluation of advanced techniques for map processing in order to streamline the cataloging of historic maps in digital libraries.

During our 12-month Tier I basic research project, we will apply automated map-processing techniques to 25 historic maps. We will compare these results with the results through two standard map cataloging methods in digital library projects. We hope to determine whether the Strabo open-source map-processing software can be used to capture the information needed to complete required fields in Qualified Dublin Core metadata records for a CONTENTdm back-end system. This is a common technical infrastructure used in many digital libraries, so our results will have broad applicability.

Media Coverage

Yao_Yi Chiang Receives NEH Grant (Media Coverage)
Publication: SSI News
Date: 12/16/2016
Abstract: Announcement of award.
URL: https://spatial.usc.edu/yao-yi-chiang-receives-neh-grant/

Unlocking the Mysteries of Historic Maps (Media Coverage)
Author(s): Dotson, William
Publication: USC Libraries News
Date: 11/8/2017
Abstract: Describes project work and preliminary results.
URL: https://libraries.usc.edu/article/unlocking-mysteries-historic-maps

Associated Products

Strabo Text Recognition Deep Learning (Computer Program)
Title: Strabo Text Recognition Deep Learning
Author: Banisetti, Sandeep
Author: Chiang, Yao-Yi
Author: Kejriwal, Lakshya
Abstract: Identifies text on maps.
Year: 2018
Primary URL: https://github.com/spatial-computing/strabo-text-recognition-deep-learning
Access Model: Open access
Source Available?: Yes

Strabo Text Recognition (Computer Program)
Title: Strabo Text Recognition
Author: Chiang, Yao-Yi
Abstract: Older version of Strabo. Strabo Text Recognition Deep learning replaced this to improve precision and recall. Strabo Text Recognition Deep learning D the source of current precision recall data.
Year: 2017
Primary URL: https://github.com/spatial-computing/strabo-command-line-pub
Access Model: Open Access
Source Available?: Yes

Source Code for the Publication Automatic Extraction of Phrase-level Map Labels from Historical Maps: (Computer Program)
Title: Source Code for the Publication Automatic Extraction of Phrase-level Map Labels from Historical Maps:
Author: Chiang, Yao-Yi
Author: Lin, Hao Wen
Abstract: Takes text found by text recognition software and extracts/
Year: 2018
Primary URL: https://github.com/spatial-computing/connecting-map-labels
Access Model: Open Access
Source Available?: Yes

Ground Truth Data for Text Recognition from Historical Maps (Database/Archive/Digital Edition)
Title: Ground Truth Data for Text Recognition from Historical Maps
Author: Chiang, Yao-Yi
Abstract: Data used for ground truth analysis.
Year: 2015
Primary URL: https://github.com/spatial-computing/map-ocr-ground-truth
Access Model: Open Access

Strabo Test System for Public Use (Web Resource)
Title: Strabo Test System for Public Use
Author: Rafique, Zahid
Abstract: Test instance where users can upload digital maps and get results from Strabo.
Year: 2018
Primary URL: http://services.digitallibrary.usc.edu/straboweb

Unlocking Maps for Discovery and Other Purposes (Public Lecture or Presentation)
Title: Unlocking Maps for Discovery and Other Purposes
Abstract: Presented results of Strabo test.
Author: Holmes-Wong, Deborah
Author: Chiang, Yao-Yi
Date: 05/15/2018
Location: Washington, DC, USA.
Primary URL: https://drive.google.com/file/d/1DdNx-koFRIl7r-pJKdxR7N_1IDc5XkZ4/view?usp=sharing

Unleashing the Power of Historical Maps (Conference Paper/Presentation)
Title: Unleashing the Power of Historical Maps
Author: Chiang, Yao-Yi
Abstract: Detailed knowledge of the natural and human-induced changes on Earth is crucial to understand human activities at the local, regional and global levels across the scale of time. Yet, the data sources that can provide such knowledge over a broad temporal or spatial scale are dispersed and not in a readily usable format for data analytic tasks. At the Spatial Sciences Institute, University of Southern California, we are building novel map processing technologies to unlock detailed geographic information from maps. Our open source software, Strabo, “reads” scanned maps for automatically identifying historical locations of places. Our recent collaboration with a title insurance company in the United Kingdom illustrates the significance of this work. Using Strabo, the insurance company is automatically reading historical Ordnance Survey maps (circa 1900 – 1970) covering the entire U.K. to identify likely locations of subterranean contamination, such as factories, mines, quarries, and gas works, which no longer exist and which otherwise would not be known today. In 2016, we were awarded by the National Science Foundation for a three-year project to build spatially and temporally linked geographic datasets from all editions of the U.S. Geological Survey (USGS) historical topographic maps. This NSF project will develop an approach and software that automatically transform the geographic information in thousands of historical maps to machine-understandable datasets. With such large-scale datasets, researchers in biology, cancer and environmental epidemiology and social sciences will have unique opportunities for answering important questions in studies that require long-term geographic data analysis.
Date: 12/01/2017
Primary URL: https://imiamaps.org/track-3-cartographic-research/
Conference Name: International Map Industry Association

Querying Historical Maps as a Unified, Structured, and Linked Spatiotemporal Source (Keynote) (Conference Paper/Presentation)
Title: Querying Historical Maps as a Unified, Structured, and Linked Spatiotemporal Source (Keynote)
Author: Chiang, Yao-Yi
Abstract: Historical spatiotemporal datasets are important for a variety of studies such as cancer and environmental epidemiology, urbanization, and landscape ecology. However, existing data sources typically contain only contemporary datasets. Historical maps hold a great deal of detailed geographic information at various times in the past. Yet, finding relevant maps is difficult and the map content are not machine readable. I envision a map processing, modeling, linking, and publishing framework that allows querying historical map collections as a unified and structured spatiotemporal source in which individual geographic phenomena (extracted from maps) are modeled with semantic descriptions and linked to other data sources (e.g., DBpedia). This framework will make it possible to efficiently study historical spatiotemporal datasets on a large scale. Realizing such a framework poses significant research challenges in multiple fields in computer science including digital map processing, data integration, and the Semantic Web technologies, and other disciplines such as spatial, earth, social, and health sciences. Tackling these challenges will not only advance research in computer science but also present a unique opportunity for interdisciplinary research.
Date: 04/06/2017
Primary URL: http://exploringoldmaps.informatik.uni-wuerzburg.de/2017/keynote.html
Conference Name: Second International Workshop on Exploring Old Maps

Unlocking Maps for Discovery and Other Purposes (Conference Paper/Presentation)
Title: Unlocking Maps for Discovery and Other Purposes
Author: Holmes-Wong, Deborah
Author: Chiang, Yao-Yi
Abstract: The USC Libraries and Spatial Sciences Institute are using Strabo software to "transcribe" text from 25 digital maps. They are evaluating the effectiveness of the software by comparing it to human transcription and OCRing of text and adding the text to metadata records for each of the maps.
Date: 10/25/2017
Primary URL: https://drive.google.com/file/d/0BzpgmJyOLhecRHZyazRCT2xybHc/view?usp=sharing

SRC: automatic extraction of phrase-level map labels from historical maps (Conference Paper/Presentation)
Title: SRC: automatic extraction of phrase-level map labels from historical maps
Author: Lin, Hao Wen
Author: Chiang, Yao-Yi
Abstract: Historical maps are important resources for various kinds of studies, providing insights for natural science and social science studies such as biology, landscape changes, and history [1]. However, text recognition on maps remains a challenging task because map usually has a complex background in which textual content appears in numerous colors, fonts, sizes, and orientations. Even if we were able to acquire perfectly recognized words and characters automatically, it is still difficult to generate useful information because individual words are not meaningful. For example, a typical result from OCR scanning or manual map digitization is that each recognized bounding box only contains a single word (Figure 1).
Date: 11/07/2017
Primary URL: https://doi.org/10.1145/3178392.3178400
Conference Name: ACM SIGSPATIAL

Linking Historical Maps to the USC Shoah Foundation Visual History Archive. In Proceedings of the 28th International Cartographic Conference (Conference Paper/Presentation)
Title: Linking Historical Maps to the USC Shoah Foundation Visual History Archive. In Proceedings of the 28th International Cartographic Conference
Author: Chiang, Yao-Yi
Abstract: Not available
Date: 07/02/2017
Primary URL: https://icaci.org/wp-content/uploads/2012/01/icc2017-program.pdf
Conference Name: 28th International Cartographic Conference

---