I’m curious to find people who are encoding coordinates (or standardized addresses) in library records for non-map materials.
I know that there are spaces within Dublin Core (http://dublincore.org/documents/dcmi-point/), MARC (255 or 034) and VRA Core for including coordinates, but I haven’t found examples of catalogers who are routinely specifying either geographic coordinates or standardized addresses for non-map materials.
A few examples:
1) A book describing the history of a particular building or intersection
2) A lithograph of a 19th century street scene
3) A photograph of a person in front of a recognizable building
3) A manuscript collection related to the history of a particular place.
Some context: I’ve recently worked with students
and “hacktivists”on pulling data out of library systems and exploring those data through various data visualization tools. We’ve found that the geographic data in many of the records is often either not included or not standardized, so that a lot of data cleaning needs to be done in order to map the contents. However, in many cases, an addresses or other standard geographic information is included in plain text notes field.
A collection of Philadelphia librarians and archives would like to learn from someone who has already established best practices for coordinate-level or address level encoding of places.
The answers fell into a few categories 1) Encouragement —————-
Quite a few people wrote saying that they had struggled with this question or variations of it, or that they had collections that would really benefit from a good answer to these questions. That was great to hear, and I appreciated those responses quite a lot.
2) Approaches based on taking advantage of the coordinates that are included in Geographic Authority files.
Some background on this approach if you’re not a librarian is at the very bottom of this post.
The two geographic authority files that I know of that are most widely used in libraries, archives, and museums are the Getty Thesaurus of Geographic Names (TGN), used very widely in visual resource description and lots of other contexts, and the Library of Congress Name Authority Files (which include Geographic names as ‘corporate names’) which is used most widely in MARC records for library catalogs.
The responses in this category tended to point out that geographic name authority records often include coordinates. This is awesome and important. However, it’s worth pointing that the TGN explicitly states that their coordinates aren’t really for mapping, and the LC Name Authority files for geographic names are not complete as far as I can tell. I really love this idea, but I don’t think I got any examples back from my query where people were using these authority records to do mapping directly. Instead, they mostly re-processed the geographic info after cataloging, as in 3 below.
3) Approaches to encoding the geographic information after the item is cataloged.
There were a few cases where I was pointed to a map of some collection, but generally when I followed up with the makers of the site, they had actually pulled the records out of the system and then geocoded them separately, through an API with lots of data cleanup, or some by hand. There’s lots of interesting work being done there, and still to be done, but it doesn’t address my particular question, which is how metadata records could be created that wouldn’t require that.
4) Using the Address or Coordinates Fields in Dublin Core or MODS
Both MODS and Dublin Core allow space for geographic coordinates directly in the records (it’s in DC-extended). I think that is awesome. It still doesn’t solve the problem of knowing what level of geographic specificity is being used so you would need a lot of rules about what coordintaes one would use. After all, real places aren’t actually points — they take up space and frequently have fuzzy boundaries. But it was gratifying to hear of a few places that were using this approach (see below for the two examples).
Overall Struggles with Geographic coordinates in records
1) Geojson would be awesome. How would a cataloger do that technically?
2) Who makes the decision about what constitutes “Delaware Valley” or a neighborhood called “East Passyunk neighborhood”
3) For images, geographic coordinates are the simplest, and even better the coordinates of the viewer if you’re cataloging a photograph or painting.
4) Addresses change over time, so you’d need current address and historic address, which I haven’t seen anyone doing. (especially for old places like Philadelphia, where the same physical building might have had a dozen addresses over its life)
Relevant Links and Resources:
- Hardy, D., & Durante, K. (2014). A Metadata Schema for Geospatial Resource Discovery Use Cases. Code4Lib Journal, (25).
- a discussion recently on how to deal with bounding coordinates
- Harvard seems to working on adding coordinates to non-cartographic materials: http://sanger.hul.harvard.edu:8080/geohollis
Ohio History Connection (formerly Ohio Historical Society) is doing with such collections as these: <http://www.ohiohistoryhost.org/ohiomemory/learning-resources/then-and-now-maps> and [http://www.remarkableohio.org/Index.aspx ](http://www.remarkableohio.org/Index.aspx)
- http://connect.ala.org/files/Cartographic_Resources_CIG_Midwinter_2014_detailed.pdf — great place to look for discussions of how MARC data has been used to make maps, and some of the trouble with encoding points data in maps. Also a good description of why and how bounding boxes and points are not the best sources for this.
- http://www.stonybrook.edu/libmap/coordinates/seriesb/no8/b8.htm – mostly about cataloging maps, but addresses some specific issues with coordinates-based approach.
Descriptions of Current Practices at two places:
Our digital repository uses MODS 3.5 which doesn’t have an element for historic or current address. There’s the cartographics subelements but this is only scale, projection, and coordinates. A lot of our digital content about Connecticut has metadata that lists the historic or current address of where the image was taken. It was necessary to keep this information. Lacking an appropriate slot, we decided on the MODS element with a type value of historic/current address. This looks like . We’ve used these for our non map materials that have any information about addresses. For our maps, the information can also have this information or is more specific to FGDC with bounding coordinates and not a physical address.
If you are using MODS, you can always use the MODS and put in any schema and information you want whether it is VRA, FGDC, DC, etc. For coordinates, I would look at the MODS list. I started a discussion recently on how to deal with bounding coordinates (http://listserv.loc.gov/cgi-bin/wa?A1=ind1410&L=mods).
For the bounding box coordinates, we map these to and use FGDC and not geoRSS and we also map to one MODS field. That way we have information in both places just in case. We selected FGDC because it is better for bounding box information and we will be migrating information from ArcGIS to our digital repository which can be exported as FGDC; this means we can store the entire FGDC record in the mods.
We have been routinely supplying coordinates for digitized visual resources (mostly photographs) for a few years now, and are in the midst of designing a public digital library portal that will allow users to browse our collections on a map, as well as by more traditional means such as subjects, genres, names, etc.
Part of the reason we began doing this is that we saw how hungry various UCLA GIS projects were for resources that had coordinates. Initially, all we could give them were maps, since the existing metadata already contained coordinates. Then we branched out to enhancing the metadata for our Los Angeles Times photograph collection, because so many of the locations were easily identified either by the photograph itself (e.g. street addresses or intersections), the existing metadata, or by a quick check of the article the photographs originally ran with (accessed via the ProQuest Los Angeles Times database). Since then, we have expanded to other digital collections as part of our standard practice whenever we are creating or enhancing metadata for digital collections.
We ask our catalogers to try to identify the spot where the photographer was standing, if possible, or an exact address for the location. If that degree of specificity is not possible, we try to identify the location within the equivalent of a block or two. If we can only identify the city, we don’t geotag that particular resource; at some point, we hope to automate the geotagging of resources that don’t have coordinates already in the metadata—although we are still exploring the best way to do that!
Currently we’re using this tool to help identify coordinates: http://itouchmap.com/latlong.html
We’re in the midst of migrating from a locally-created DAMS to Islandora, so this metadata is stored in a variety of ways now, but eventually we are planning to map it all to MODS.
Report from the Best Practices in Geospatial metadata at DLF.
Notes for the non-librarian on Authority files:
(If you’re a librarian and can suggest ways I can make this simpler or clearer or more accurate, I’d appreciate it)
“authority files” are closely kept lists of terms that serve as the Authority for catalogers who create metadata. An over-simplified (and possibly somewhat inaccurate) example follows. When she published something that needed cataloging, a name authority file was created for danah boyd. It includes her birth data, and is linked to another variation of her name. That way, when further works of hers are cataloged, the catalogers will know exactly how to spell her name, and the appropriate “see” references will make their way into local catalogs for her.
In general, cataloging practices for a particular collection will describe the metadata standards that will be used (Dublin Core, MARC, MODS, VRA Core, etc), as well as the authority files to which the catalogers should look to get the names.