Welcome back to my blog! I took my break day a week early for this class due to the conference that I mentioned in my last blog, but I’m excited to be back and working through the Text Analysis material for our Digital Methods course!
While I have used some Text Analysis procedures before, it was really helpful to go through our introductory readings, as they were able to provide a nice overview of the different things it can actually do. When I did my own personal crash course, I started with the knowledge that I should be able to do things A, B, and C – I just had to learn how. Therefore, I never really got an overview of what Text Analysis can be used for and is used for across and among researchers, especially those in the humanities who are pursuing projects more related to what I might explore for part of my dissertation or as part of Campus Archaeology. Ted Underwood’s (2015) piece titled Seven Ways Humanists Are Using Computers to Understand Text was extremely helpful in this sense, as it highlighted some things I’ve done before, such as compiling word counts and word clouds, but also introduced the idea of and potential problems associated with unsupervised modeling. As this is something I was interested in exploring in future research looking at journal articles to see if I could classify articles into subjects – and if/when that didn’t work, I could explain why we might be seeing so much overlap within biological anthropology despite the divide between forensic anthropology and the rest of the subdiscipline – I really valued Underwood’s (2015) advice to think about what types of assumptions such a model might be making and how that affects any results. Even if I had an idea that such a model might not work, what does “working” really mean when I’m personally assigning categories to articles or data, or when the model is working with assumptions we as researchers would not usually make within the field of biological anthropology.
However, in particular, I found Lisa Rhody’s (2015) The Story of Stopwrods: Topic Modeling an Ekphrastic Tradition and Heather Froehlich’s Moby Dick is About Whales, or Why Should We Count Words to be incredibly thought provoking. And the reason for that is although these authors tackled two “easier” factors of Text Analysis, they both broke down the different assumptions we might be making if we use these methods without thinking.
Lisa Rhody’s (2015) The Story of Stopwords: Topic Modeling an Ekphrastic Tradition touches on the removal of stopwords in Text Analysis and whether it is actually appropriate to take them out of an analysis. While common words such as “and”, “so” or “because” may seem irrelevant, they may actually be extremely insightful transitions or word choices that we are completely obliterating from an analysis. As per my last blog, my recent conference presentation focused on the composition of forensic versus non forensic articles in the American Journal of Physical Anthropology (AJPA). And during this project, I removed all standard stopwords, as well as several of my own that were provided by the journal as categorical descriptors (e.g., “Abstract”, “Methods”, “Materials”, “Results”, etc.). While I was able to rationalize the words that I personally took out, I can’t say I took the time to look at every word included in the stopwords R package… And looking back, this may have very well affected my analysis! I didn’t even consider the possibility that commonly considered “simple” words may be used in a different way within research articles. Although this may not have greatly affected my project, I now know how important it is to think about every aspect of the choices I’m making – and to remember that I don’t have to make the same choices as another study just because it worked for them.
Heather Froehlich’s Moby Dick is About Whales, or Why Should We Count Words really made me think about the possibilities involved with approaching Text Analysis. In this piece, Froehlich shows that word count alone is only so informative; however, it can also be used to understand how very common or even rare words are used in context, such as how synonyms might be used for different purposes (e.g., ship versus boat). In the project I introduced in my last blog, some of my results were visualized in word clouds and, as one example, showed that “age” was a very common word in both forensic and non forensic articles in the AJPA. However, I did not look at the context of how “age” was used in each type of article – in fact, I just assumed they were the same thing! Likely due to my bias as a forensic anthropologist in training or even my bias as a developing researcher (aka a student). For reader context, forensic anthropologists commonly create methods to estimate the age of unknown decedents in order to aid in their identifications for medicolegal investigations. Therefore, age often refers to these specific methods regarding their creation and validation. However, when thinking about the context of the overall field of biological anthropology, age could be used to describe the actual age of individuals (or even primates) in a study or even the age of a site or fossil! Luckily, while this may not have been included in my conference poster, this way of thinking about my own work has opened new doors and areas of exploration within my project and may help me expand it in more meaningful ways down the line.
All in all, I love reading through these articles. While I’m seeing that I may not have done everything right in my project and/or have areas to expand upon, I know that this just makes me a more competent researcher in the long run. Methods aren’t just black boxes where input A mysteriously becomes output B – methods are tools which we can use in a variety of ways if we take the time to understand them properly. And I can promise that I’ll be doing more and more of that in the future!
