3/22/2022 - A Brief Discourse on Copyright

The purpose of this post is to ask some legal questions before delving into the technical aspects of my next project. I want to make it clear that this is an opinion piece which asks many questions and offers very few answers, while remaining purely hypothetical. In the past, I had an interest in Early American Gothics which meant that any texts I wanted to use were in the public domain, simply due to their date of publication. This meant that I could do pretty much whatever I wanted with these texts, including digitizing them into plain-text copies that I could analyze (of course, the challenge with these older texts was the actual digitization since accessible copies are sometimes poorly preserved!).

However, let's imagine another scenario: I want to, hypothetically, digitize all of Stephen King's horror novels that are sitting on my shlef in order to analyze word choice changes and sentiment changes over the course of his career, and perhaps compare that chronology to contemporary horror movies/novels that were published in that time. In order to do this, I would need plain-text copies of his novels, which are not in the public domain, and would therefore beg the question of what we are allowed to do with them due to copyright.

Generally speaking, if you are doing academic work and you are not making money by using a text (or other media outside of our little scenario), you can usually claim 'fair use.' Yet fair use typically means pulling a quote or other small excerpt, rather than using every single word from a given text. I'll go into detail later about what factors decide fair use after we have a little bit of fun.

Let's continue our hypothetical: I go ahead and scan all of Stephen King's novel using some sophisticated system that turns his physical books into perfect plain text documents that we can conduct various NLP analyses on. In the subsequent paper that I write, I use these results to make some kind of insightful claim about King's writing style. In good practice, I post a repository of my code and list of texts used, as well as details on how certain graphs/figures were produced. Do I provide a repository of the plain-text copies of the novels I used?

On the one hand, reproducability is essential to the scientific process and the more information available, the more it is reproducible and generally experimentally sound. On the other hand, we are providing the entire text of multiple copyrighted novels, which someone could freely access and then use for non-academic purposes. Does this repository still fall under fair use? I would imagine that it would not. Would the same project, but without a public repository of texts available be fair use? This is a little bit of grey area. I bought all of the books, which means I should be able to do anything I want with these physical copies, including scanning them and digitizing them (or burning them if I so chose), as long as I'm not then sharing those scans with anyone. On the other hand, I am making a complete copy of the text of the novels, which does not feel like fair use when worded so generally, but suddenly feels fair use if I switch out 'text' with 'data.' This feels even more confusing if I were to choose 'content' instead of 'text' or 'data' since the question appears: does data still contain meaningful content?

Confused yet? Let's expand this hypothetical further: suppose I am a professor (perhaps one day!) and a colleague of mine in the same department wants to do a follow up analysis to my Stephen King project because they found it to be so insightful! Because I own all of the King novels and because the plain-text I derived from them shouldn't be shared in order to not violate copyright concerns, I ought to not share these plain texts because my colleague could theoretically leak these texts which would mean that my use of the texts is probably no longer fair use. Does this feel silly? Does it feel a little bit like being labelled a criminal for sharing your netflix password with a close friend? I suppose if I trusted my colleague to not leak these texts, they could claim to have performed the exact same process that I have in order to end up with identical plain text copies, and no one would ever be able to know if indeed this claim is false.

Well we need not lose too much sleep over all of these questions and moral quandries--there is some precedence for the very question of whether text data from books falls under fair use. In particular, there are two court cases that are useful: Authors Guild, Inc. v. Google, Inc. and Authors Guild, Inc. v. HathiTrust. In summation, Google began to digitize texts by scanning them and making what was essentially a virtual library where users could view a given book online. Due to OCR errors being likely, scanned pages were available for users to view instead of plain-text (this allows users to self-verify). Initially, Google only used books in the public domain, so there weren't really any problems, but then they established library partnerships with various institutions and began the process for books that had copyrights. Publishers obviously weren't too happy and a lawsuit ensued. Through a lengthy process of settlements and appeals, it was agreed upon that the Google Books program fell under fair use, though many questions remain. This ruling was the basis for controlled digital lending (CDL), which suggests that entities could 'lend' digital copies of a book much like a physical library could. However, this concept is currently being testing in court (Hachette v. Internet Archive) so we don't know if there is a legal basis for this concept.

Obviously, this scale is much larger and more complex than the questions we are dealing with in our hypothetical, but these cases are closely related to our problem. Ultimately, we should ask ourselves whether this process violates the four factors in #17 U.S. Code § 107: 1. The purpose and character of the use, including whether such use is of commercial nature or is for nonprofit educational purposes. 2. The nature of the copyrighted work. 3. The amount and sustainability of the portion used in relation to the copyrighted work as a whole. 4. the effect of the use upon the potential market for or value of the copyrighted work. Utilizing these four factors, we can estimate whether our process falls under fair use.

1. We are using this for nonprofit and educational purposes. We should be safe here.
2. We are not using these texts to copy any themes or other aspects of King's work. We are probably safe here too.
3. We are using the entire text rather than an excerpt. We could be in trouble here, but if we are only sharing graphs and other conclusions based on our analysis, we should probably be fine, but there could be grounds for argument here--it would mostly depend on what we actually share externally (think back to the quesiton of sharing the plain text repository).
4. Unless our paper becomes very famous and actively changes the sales of King's books, this should not be a problem either. Odds are, sales would only increase! However, in the event that we negatively changed the sales of King's books, an argument could be made that our final paper is the same as a criticism of the most recent Broadway play that you might read in a newspaper. I think we're probably safe on this one too.

Utilizing these factors, I think we can feel safer about creating plain text copies of these copyrighted works for our analyses, as long as we are not publicly sharing these files. For the issue of sharing your files with your colleagues, I will let you ponder whether you would trust them with your Netflix password...I know I wouldn't trust any of my colleagues! More importantly, at the end of the day we should ask ourselves if Stephen King (or more specifically, his publishers) would actually sue us for such an endeavor. I'd like to think not! I hope that this discourse helps us to think through the ramifications that our work as researchers might have beyond our actual research itself, or is at least somewhat informative. I also think that it is good practice to consider such things before embarking on potential research projects; we'll see if my OCR project is successful or not!