Tag Archives: Text Mining

Voyant, CartoDB and Palladio: A Comparison

Using the same data for all three project was an enlightening experience. Intellectually I had accepted that different methodologies would yield varying results and elicit different questions. But actually doing to work has given me a deeper understanding of the magnitude of this concept.

As an archaeologist, I’m always very aware of place. I often ask “where” questions and think of  human activity taking place on the crust of the Earth at specific points. Therefore, CartoDB was fairly intuitive to me. Before beginning to play with the software I already had some idea of what producing a map could tell me. What did surprise me was the depth of information I was able to get about place with the other two tools.

Although I often work with texts to both point me in the direction of sites and to add nuances to the archaeological record, I’ve always struggled conceptualizing text within space. Voyant in conjunction with Carto helped me visualize this relationship. While Voyant gave me visualizations about words that occurred within the seventeen states, Carto helped me make spatial connections and situate these data in a place rather than just a word “Alabama.”

Likewise, Palladio helped me make further connections about the observations I had made in Voyant and Carto. Voyant acted more as a comparative tool. I could see how word frequencies changed across the corpus. Palladio was a comparative tool as well, but graphs visualized magnitude and categories, whereas Voyant was useful in discovering that these categories existed, but was less effective in presenting observations in relation to other data.

The observation I made after looking at the data in all three tools was that there was a significant movement of people after emancipation. Voyant provided the words used to describe this movement like place names and occupations following freedom. CartoDB conveyed how far afield people traveled after emancipation. Finally, Palladio showed me the movement of individuals. That dynamic action across time and space is not something that one application was able to fully convey.

That being said, the Voyant, CartoDB and Palladio each have their specific strengths. Palladio might have a mapping feature, but if your project has a heavy map component, use Carto. Voyant can be used to topic model, but use Palladio to visualize how the topics related to people. Carto can insinuate relationships, but rely upon Palladio to actually connect the dots.

After looking at the three tools side by side I can see real potential for projects that integrate more than one. That being said, I find that academics can get lost in the sea of knowledge. Some times we spend so much time trying to know everything we can about a topic we loose sight of our project. A successful project needs to be able recognize when a tool will be useful and when it will detract from the goal. These three programs are very powerful discovery and publication tools. I find it very challenging to balance discovery with putting knowledge out there. At some point I have to at least pause discovery, draw conclusions, and share what I’ve learned. And sometimes I find it incredibly fruitful to return to the discovery process. Palladio and CartoDB allow for that fluidity, whereas Voyant is much harder to return to.