By Estelle Jung
Topic Modeling Reflection
For my topic modeling assignment, 128 documents were used in total. The documents were from the two volumes of The Spanish Regime in Missouri by Louis Houck. At first, I tried to use 20 topics, but Mallet gave me the same weight of 0.25 for each topic. After changing the number of topics to 15, I ended up with various weights which is what we wanted. Thus, 15 seems to be the optimal number of topics for this example.
I think that there were two coherent topics apparent in this assignment based on the data from the Keys spreadsheet. Topic 13 had the highest weight of 0.69804. The top words of topic 13 included “order,” “made,” “time,” “great,” “make,” “part,” “lordship,” “men,” “letter,” “government,” “indians,” “good,” “present,” “give,” and “governor” which suggests that the words surrounding this topic related to the relationships that were developing between the Spanish and the Natives. For instance, words like “government,” “governor,” and “order” suggest that the Spanish were established authority and dictated the rules for the “indians” through the exchange of “present(s)” and correspondence of officials with “letter(s).” Moreover, the top words “made” and “make” imply that the Natives were forced to succumb to the hands of the Spanish. Furthermore, it is important to note that Topic 4 was a prominent topic in this data spreadsheet with a weight of 0.65304. The terms “don,” “luis,” “san,” “years,” “order,” “royal,” “post,” “general,” “pedro,” “village,” “carlos,” “Louisiana,” “lordship,” “province,” “captain,” and “grace” were present in this topic. It can be concluded that these top words such as “general,” “royal,” and “don,” have some meaning attached to the large influence of Spanish military and governing officials. In other words, both Topic 13 and Topic 4 address the prevalence of Spanish rule over the Natives in the Upper Louisiana region. This theme is seen through the selection of top words that describe the names or titles of Spanish leaders and words that depict the interactions between both groups of people.
Unfortunately, I was unable to turn my topic modeling data into a visual representation on Tableau due to technical difficulties. The issue was that my composition and keys file would not join together based on the Topic ID. I have both the composition file and the keys file formatted correctly, but when I upload both documents onto Tableau, they are unable to merge together into one document. If this problem could be resolved, I would have a graphical representation of which topics stand out apart from others. I believe that a good topic model is definitely possible for this set of documents. In fact, topic modeling will be necessary for answering my research questions for the final project because it is a type of text-mining tool. This method reveals a group of words that best represent the information or background of a collection of documents. Topic modeling makes understanding the bulk of two volumes easier because it breaks down the text into overarching themes. Along with the other alternative methodologies like Recognito, Gephi, and Voyant Tools, Mallet and Tableau offer a different approach in which the user has access to identifying the key points in a considerable amount of text. Personally, I want to explore the advantages of topic modeling for the final project because I will be able to tell what the main argument or focus of the author is based on the top words of each topic.
In short, I have to fix the error in Tableau so then I can move on to the next step of transforming my topic modeling data into a visual diagram for better analysis. Breaking down both volumes of my primary source into multiple topics will help other students and historians understand the intentions of the Spanish in the Upper Louisiana region. By examining the motives of the Spanish, we can then investigate religious, political, social, and cultural interactions between the Spanish and the Natives – the driving forces behind a larger historical narrative of conquest.