Many cultural institutions have accelerated the development of their digital collections and data sets by allowing citizen volunteers to help with the millions of crucial tasks that archivists, scientists, librarians and curators face. One of the ways institutions are addressing these challenges is through crowdsourcing.
In this post, I’ll look at a few sample crowdsourcing projects from libraries and archives in the U.S. and around the world. This is strictly a general overview. For more detailed information, follow the linked examples or search online for crowdsourcing platforms, tools or infrastructures.
In general, volunteers help with:
- analyzing images, creating tags and metadata, and subtitling videos
- transcribing documents and correcting OCR text
- identifying geographic locations, aligning/rectifying historical maps with present locations, and adding geospatial coordinates
- classifying data, cross-referencing data, researching historic weather, and monitoring and tracking dynamic activities
The Library of Congress utilizes public input for its Flickr project. Visitors analyze and comment on the images in the Library’s general Flickr collection of over 20,000 images and the Library’s Flickr “Civil War Faces” collection. “We make catalog corrections and enhancements based on comments that users contribute,” said Phil Michel, digital conversion coordinator at the Library.
In another type of image analysis, Cancer Research UK’s Cellslider project invites volunteers to analyze and categorize cancer cell cores. Volunteers are not required to have a background in biology or medicine for the simple tasks. They are shown what visual elements to look for and instructed on how to categorize into the webpage what they see. Cancer Research UK states on its Web site that, as of the original publication of this story, 2,571,751 images have been analyzed.
Both of the examples above use descriptive metadata or tagging, which helps make the images more findable by means of the specific keywords associated with — and mapped to — the images.
The British National Archives runs a project, titled “Operation War Diary,” in which volunteers help tag and categorize diaries of World War I British soldiers. The tags are fixed in a controlled vocabulary list, a menu from which volunteers can select keywords, which helps avoid the typographical variations and errors that may occur when a crowd of individuals freely type their text in.
The New York Public Library’s “Community Oral History Project” makes oral history videos searchable by means of topic markers tagged into the slider bar by volunteers; the tags map to time codes in the video. So, for example, instead of sitting through a one-hour interview to find a specific topic, you can click on the tag — as you would select from a menu — and jump to that tagged topic in the video.
The National Archives and Records Administration offers a range of crowdsourcing projects on its Citizen Archivist Dashboard. Volunteers can tag records and subtitle videos to be used for closed captions; they can even translate and subtitle non-English videos into English subtitles. One NARA project enables volunteers to transcribe handwritten old ship’s logs that, among other things, contain weather information for each daily entry. Such historic weather data is an invaluable addition to the growing body of data in climate-change research.
Transcription is one of the most in-demand crowdsourcing tasks. In the Smithsonian’s Transcription Center, volunteers can select transcription projects from at least 10 of the Smithsonian’s 19 museums and archives. The source material consists of handwritten field notes, diaries, botanical specimen sheets, sketches with handwritten notations and more. Transcribers read the handwriting and type into the Web page what they think the handwriting says. The Smithsonian staff then runs the data through a quality control process before they finally accept it. In all, the process comprises three steps:
- The volunteer types the transcription into the web page.
- Another set of registered users compares the transcriptions with the handwritten scans.
- Smithsonian staff or trained volunteers review and have final approval over the transcription.
Notable transcription projects from other institutions are the British Library’s Card Catalogue project, Europeana’s World War I documents, the Massachusetts Historical Society’s “The Diaries of John Quincy Adams,” The University of Delaware’s, “Colored Conventions,” The University of Iowa’s “DIY History,” and the Australian Museum’s Atlas of Living Australia.
Optical Character Recognition is the process of taking text that has been scanned into solid images — sort of a photograph of text — and machine-transforming that text image into text characters and words that can be searched. The process often generates incomplete or mangled text. OCR is often a “best guess” by the software and hardware. Institutions ask for help comparing the source text image with its OCR text-character results and hand-correcting the mistakes.
Newspapers comprise much of the source material. The Library of Virginia, The Cambridge Public Library, and the California Digital Newspaper collection are a sampling of OCR-correction sites. Examples outside of the U.S. include the National Library of Australia and the National Library of Finland.
The New York Public Library was featured in the news a few years ago for the overwhelming number of people who volunteered to help with its “What’s on the Menu” crowdsourcing transcription project, where the NYPL asked volunteers to review a collection of scanned historic menus and type the menu contents into a browser form.
NYPL Labs has gotten even more creative with map-oriented projects. With “Building Inspector” (whose peppy motto is, “Kill time. Make history.”), it reaches out to citizen cartographers to review scans of very old insurance maps and identify each building — lot by lot, block by block — by its construction material, its address and its spatial footprint; in an OCR-like twist, volunteers are also asked to note the name of the then-existent business that is hand written on the old city map (e.g. MacNeil’s Blacksmith, The Derby Emporium). Given the population density of New York, and the propensity of most of its citizens to walk almost everywhere, there’s a potential for millions of eyes to look for this information in their daily environment, and go home and record it in the NYPL databases.
Volunteers can also user the NYPL Map Warper to rectify the alignment differences between contemporary maps and digitized historic maps. The British Library has a similar map-rectification crowdsourcing project called Georeferencer. Volunteers are asked to rectify maps scanned from 17th-, 18th- and 19th-century European books. In the course of the project, maps get geospatially enabled and become accessible and searchable through Old Maps Online.
Citizen Science projects range from the cellular level to the astronomical level. The Audubon Society’s Christmas Bird Count asks volunteers to go outside and report on what birds they see. The data goes toward tracking the migratory patterns of bird species.
Geo-Wiki is an international platform that crowdsources monitoring of the earth’s environment. Volunteers give feedback about spatial information overlaid on satellite imagery or they can contribute new data.
Gamification makes a game out of potentially tedious tasks. Malariaspot, from the Universidad Politécnica de Madrid, makes a game of identifying the parasites that lead to malaria. Their Web site states, “The analysis of all the games played will allow us to learn (a) how fast and accurate is the parasite counting of non-expert microscopy players, (b) how to combine the analysis of different players to obtain accurate results as good as the ones provided by expert microscopists.”
Carnegie Melon and Stanford collaboratively developed, EteRNA, a game where users play with puzzles to design RNA sequences that fold up into a target shapes and contribute to a large-scale library of synthetic RNA designs. MIT’s “Eyewire” uses gamification to get players to help map the brain. MIT’s “NanoDoc” enables game players to design new nanoparticle strategies towards the treatment of cancer. The University of Washington’s Center for Game Science offers “Nanocrafter,” a synthetic biology game, which enables players to use pieces of DNA to create new inventions. “Purposeful Gaming,” from the Biodiversity Heritage Library, is a gamified method of cleaning up sloppy OCR. Harvard uses the data from its “Test My Brain” game to test scientific theories about the way the brain works.
Crowdsourcing enables institutions to tap vast resources of volunteer labor, to gather and process information faster than ever, despite the daunting volume of raw data and limitations of in-house resources. Sometimes the volunteers’ work goes directly into a relational database that maps to target digital objects and sometimes the work resides somewhere until a human can review it and accept or reject it. The process requires institutions to trust “outsiders” — average people, citizen archivists, historians, hobbyists. If a project is well-structured and the user instructions are clear and simple, there is little reason for institutions to not ask the general public for help. It’s a collaborative partnership that benefits everyone.