Parker Commentary

Parker Commentary

Anonymous

Fall 2017

Matthew Parker

For my experimental bibliography, I decided to do computational text analysis of my novel. I started out by researching the methods and libraries to use for this, decided to use the NLTK and Gensim to perform latent Dirichlet allocation after researching different methods for textual analysis, and then coded a functional model using a small toy corpus and sample query text. The model produces the topic from the corpus most closely associated with a query text. Problems I encountered doing this were as follows: 1) To do computational text analysis, I needed a plaintext copy of my novel, and none currently existed. This major issue was somewhat mitigated by the existence of a plaintext copy of another novel by the same (anonymous) author. While the OCR isn't the best, I was able to obtain a relatively similar plaintext sample from Google books which I used in place of the original novel. Because of the computational complexity of the operations involved, the operation over the corpus is still in progress. Code can be found here.

Modern NLP libraries were used to analyze historical works to identify trends in a given text compared to a large corpus