Tai Xiang's Random Projects

The paintings on the back of an illustrated man, the sweet dreams of replaceable nobodies high on soma, the torturous reformations of a violent society: Science fiction has always been my favorite genre to read. It pieces together speculative visions with political commentary; builds worlds of surveillance and mystery.

I feel the themes and messages of science fiction succeed in part due to their lack of directness. Works do not state a political point to the reader, do not spout social commentary plainly in text. The reader must ease into the messages buried within scenes of drama and fantasy. Science fiction’s ideologies are always obscured, whether it be in the underbellies of farm animals or in the politics of multi-planet settlements. It feeds off those ideas of political controversy and turmoil.

I’ve always seen science fiction books, with their entanglement with political and social issues, as representation of the times they were published in. The works of an author are an outcrop of the world they grow up in, their characters and settings a reaction to the issues of race, gender, class, and war they were born into.

That bit of speculation forms the base of my hypothesis for this independent study. Perhaps there is a way to link the relevancy of historical context to the writing of science fiction texts.

I hope to seek that out through the use of sentiment analysis, a computational technique that allows one to label the states and attitudes of language and text. Sentiment analysis can be utilized to classify a text in a variety of ways including positive/negative splits, states of emotions, etc.

My goal is to create various data sets of best selling science fiction tests centered within consistent incremented sets of time in history (a increment time of ten years would center texts from 1940-1950, 1950-1960, and so on). Sentiment analysis will then be performed on these data sets and I will attempt to contextualize the results of this analysis within the historical happenings of each time period.

With the rapid development and emergence of these data science tools, the uncovering of patterns and consistencies within literature has become far more accessible. We no longer need to manually sift through massive volumes to gleam little bits of information and piece together tiny pieces of information in an attempt to come to anything remotely conclusive. Hopefully the applications of these techniques to this project will be able to reveal something insightful to both you and I.