For this project, I worked with three translations of The Library of Babel. The first two were published almost concurrently with each other. James Irby’s translation was published in 1962 for a collection of Borges’ works titled Labyrinths. Anthony Kerrigan released his the same year, for his translation of Ficciones as a whole. The third translation I worked with, was published by Andrew Hurley in 1999 for his collection of Borges’ works titled Collected Fictions.
Four techniques and programs were applied to each translation for comparison; traditional close reading, SameDiff, JuxtaCommons, and Sentiment Analysis.
Close reading was the starting point; I sat, read, annotated, and took notes on each translation. This gave me a ideas on how the translations differed, and gave me ideas on what to look for with the more technical comparisons. I avoided using information gleaned from these close readings in my argument itself, as these aren’t quantifiable.
SameDiff is an online tool found at: https://databasic.io/en/samediff/ . It allows you to upload two texts, and it will display some useful statistics such as word count. It also graphically displays the sets of words: Words only found in Text 1, words only found in Text 2, and words found in both texts. The words are larger or smaller depending on their frequency in the texts. The most powerful feature of this tool though, is it allows you to export a csv file containing a list of all the words found in the texts, and their frequency in each texts. This basically tells you the exact same thing the graphical display on the site does, but it also gives you quantifiable data. I plugged all the texts into the tool, and got three csv files for comparison - Irby vs Hurly, Irby vs Kerrigan, Hurly vs Kerrigan. I use data from these tables in my argument to quantify differences.
JuxtaCommons was the second tool I used. It can be found at: http://juxtacommons.org . At a basic level, it allows you to upload text documents and compare differences between them. Once you upload a text, the tool formats it into a “witness”, and from there you can create a comparison set of as many “witnesses” as you want. I uploaded all three texts to the tool, and used it to find syntactical differences between the texts. It was here that I realized how similar the texts are; the syntax is only slightly rearranged between them for the most part. JuxtaCommons highlights every difference in the text and also gives you a “change index”.
Using this data without interpretation is flawed though. For example, the lowest change index was .44. This would lead one to believe the texts are very different. The only reason the index is so high though, is that the tool highlights every difference, no matter how miniscule. Every individual word and every individual shifted clause increases this index.
This flaw lead me to explore Sentiment Analysis using R. By using Sentiment Analysis, I was able to quantify sentiment and emotion in each translation, to find that differences are miniscule between them. There are slight differences, but the trend lines of sentiment in each translation are all similar.~~~~~~