As an entry point into my poetics (a productive synthesis of my instructions from the three texts used for our experiment) I wanted to test the generative potential of some topic modeling software (MALLET) that has become somewhat popular amongst digital humanists. Essentially, MALLET organizes words from a set of document into a preset number of “topics” based on their statistical likelihood of appearing in the same document.
After running my 15 blog posts through MALLET, it generated the two data sets below. In the first table, the right column (Terms) is the set of terms that MALLET deemed statistically correlative, and the left column (Possible Topics) is my best guess at the conceptual string that connects these terms. It might be useful to revise the left column as I continue with my poetics or take suggestions for alternate topic possibilities in class.
relationship desire painting art position viewer importance material artistic individual attention locating things determined focused artist social author extrapolating
images essay moving process filmic creation experiment audio representation felt unique bresson event metz object collection put logic project
writes photography movement frame role kind world campany terms absence late elements part don control means mountain element doesnt
media julien search bragaglia specific focus tension claims locate photodynamism technology poetry poetics list initiated earlier subtle explicitly explores
subject real agency human lecture approach freud repetition theory network act moments encounter consciousness leads relation traumatic place back
jullien shi contrast metaphysics chinese general western ulmer reality move thinking eastern situation culture aesthetic dynamism victory agent disposition
lacan unconscious function signifiers language conscious relations essential signifier description identical system pre constitutive mapping cut processes state understanding
electrate important term metaphysics method concept century level matter characteristics word fact due definition question place made describes addition
image time film point present de duve nature refers photograph motion simply difficult order past photographic barthes literacy psychoanalysis
instruction notion thing post space text moment section mind formulating contradiction instructions structure subjective clear similar posits claim binary
Most of the topics are fairly predictable as is their distribution which essentially aligns with the three different texts. In coming up with names for my topics, I just read the list of words two or three times and wrote down the word that came to my mind. I think the best topics I got were “Precision” and “Experimentation,” a somewhat contradictory pair, both of which figure prominently in my cinematic blogs. Topics such as “Unconscious” and “Psychoanalysis” were not very illuminating, although I think its interesting that “interpretation” figures prominently in my lacan posts.
Topic Prevalence 1
Topic Prevalence 2
Topic Prevalence 3
Of course, this data is only really valuable for its generative potential considering it is almost entirely dependent on interpretation. But, since “all roads lead to Rome” (i.e. the unconscious), I don’t see a reason why we can’t use tools like this in electracy for mapping constellations of signifiers. Maybe an instruction could be formulated from the most prevalent topics in the three blog sets (unconscious, fluidity, and experience). Here is an attempt: An electrate metaphysics explores the unconscious fluidity of experience.