Yesterday I tried out the “Serendip-o-matic,” a new research tool developed at the Roy Rosenzweig Center for History and New Media. Using textual analysis, this program takes the words you put into it to find relevant sources from places like the Digital Public Library of America, Europeana, Trove Australia, and Flickr Commons.
When I first tried it, I wasn’t particularly impressed with the results. After reading Liz Miron’s post on this subject, I was inspired to try her method of pasting an entire article or paper into the search. It hadn’t occurred to me to try that. I was still in the mindset of Google searching, where only a few keywords are necessary. But let’s get back to the Serendip-o-matic.
My findings:Medieval Cartography
- A search of “cartographic, cartography, map, mappaemundi, mappamundi, medieval” yielded 4 sources. All of these were related to cartography, but only one of them might have been relevant to my paper (a bibliography of works relating to cartography in the Hathi Trust). None of them were medieval.
- A search of my entire paper on medieval cartography yielded 36 sources. Some of these were not at all related to my subject (for example, “Map of the Murrumbidgee Irrigation Area” from Flickr), but at least one was an important source that I referred to time and again. Several of them were medieval.
- A search of “factory, girls, industry, lowell, mill, women” yielded 28 sources. Most of them were photographs. Two of them were directly related to my paper on the subject of female factory workers in Lowell during the 1800s.
- A search of my entire paper on the Lowell mill girls yielded 36 sources. Many of them were photographs, especially photographs of factory workers on strike. None of them were directly related to my paper.
- A search of “America, cold, war (the Serendip-o-matic split up my search of “Cold War”), communism, containment, fear, film, stereotypes” yielded 28 sources. Many of them were photographs. There were a few about the American film industry, but nothing relevant to my paper on the portrayal of communism in 1950s American Sci-Fi.
- A search of my entire paper on this subject yielded 4 sources. All of them were about the film industry, but none of them were relevant to my research.
Perhaps the Serendip-o-matic should change its name to “the hit-and-miss.” I noticed that when I pasted entire papers into the search, Serendip-o-matic took frequently occurring keywords out of them and searched for those keywords. Makes sense, but I think there’s a flaw in the program when one of the keywords it searches for is “also.” (This happened in the 1950s film search.) I didn’t like how it split up phrases like “Cold War” into separate search terms. The sources that it did pull up were usually related to my research only in a very general way; and much of the time, they weren’t relevant at all.
That being said, it is a great idea. With the number of databases and digitized collections being created, it would be nice if there was somewhere for historians to do a one-stop search. Textual analysis also holds a lot of promise for researchers in the twenty-first century. Distance reading is a promising way to reasonably access the multitude of sources now available to us (the embargoed sources in the Hathi Trust can only be accessed this way). And for a program that was just launched a month ago, Serendip-o-matic does fairly well. But it needs improvement. At this point, a Google search seems to yield better sources most of the time.
So how can we make projects like Serendip-o-matic more effective? An important place to start is the number and diversity of records in the database. I think that this tool would be better if it also searched through collections in (for example), the Internet Archive, Project Gutenberg, the British Library, and so on. But then again, as other projects like the Internet Archive digitize more primary sources and unpublished material, perhaps the Serendip-o-matic will be quickly rendered obsolete.
Conclusion: I’ll definitely be keeping an eye on this program, but I don’t think I’ll be using the Serendip-o-matic for my research anytime soon.