Comparing Voyant, kepler, Palladio
A conventional comparison between the visualizations offered on Voyant, kepler.gl, and Palladio sticks to their strengths. Voyant excels in text analysis, kepler in geospatial analysis, and Palladio in networks analysis. But the sites and attendant softwares are not as separate as they seem. Palladio’s principal interface is a map, after all, and Voyant’s “links” and “WordTree” draw networks. It is reasonable to ask, therefore, whether analyses of words, places, and relationships can occur simultaneously on one of these sites, rather than separately on three different interfaces. The answer is that not all maps are the same, and that consequently, digital humanities sites may be better off combining elements of Voyant, kepler.gl, and Palladio, than opting for one or the other.
A more refined comparison of the three sites reveals a clear pattern. On Voyant, we can hunt for needles in haystacks—the needle, in this case, a single word or phrase, and the haystack a text or corpus. We can measure the phrase’s ebb and flow across time, and its association with other words in context. But whatever tool we apply, we are looking at a tree and not the forest. On kepler.gl, the same applies to space. We are looking for a discrete phenomenon or class of events, whether interviews or tornadoes, and we can track them not only in space, with kepler’s “trips” tool, but also in time, with its timelines. With Palladio, the pattern ends in a synthesis of Voyant’s textual and kepler’s spatial worlds in Palladio’s “maps” and “graphs” functions.
Maps: But the distinctions between kepler.gl’s and Palladio’s map functions soon become apparent. On Palladio, the map serves two purposes: to mark points or relationships between points. What it does not provide are the full metadata that lie behind a given point: the date of the interview, the place, the age of the interviewer, and so forth. I say “full” because a particular layer of metadata does appear, as chosen by the map builder. But the beauty of kepler.gl is that when we brush over a point, a pop-up window shows us everything that it has to tell, with each layer or metadatum labeled clearly: Birth–July 25, 1960. Kepler.gl also offers us a broader palette of graphics with which to visualize the points’ relative significance than Palladio does. Thus, where Palladio measures relative size through the size of a dot, kepler.gl also measures it through “density,” in color. And what about networks? Here, the Uber template also has an advantage over Stanford’s Palladio. Palladio’s lines between points are crisp and attractive. But kepler.gl allows us to choose between lines and arcs.
Graphs: Palladio’s word network graph is in my view more visually impressive than Voyant’s. But the graphs’ capacities and utilities are no different. Thus, if forced to throw one overboard and draw on just two web apps for our DH visualizations, we would be best advised to heave Palladio in favor of Voyant and kepler.gl. If asked to map Mark Twain’s worldwide lecture tour of the 1890s, when his investments in a friend’s invention tanked and he left Connecticut for cheaper climes in Paris(!), we would draw on kepler to track Twain’s “trips,” and intersperse images from Voyant to see where the trips appeared in Twain’s journalism. We could limit the results temporally, by showing graphs of Twain’s usage of “Italy” in spring 1896, after he passed through there.