1

I've been working on building a Dr. Seuss dataset which provides training data for my recurrent neural network. The goal is to use machine learning to get the computer to generate Seuss-like works.

I was looking for an answer online and I saw that Seuss stuff isn't going to be in the public domain for quite awhile. But, does hosting it on github for educational/open source purposes put it under fair use?

To be clear, it's just a bunch of plaintext files with Seuss books/poems in them. They aren't all perfect, and of course there are no illustrations, and the 'form' is rarely perfectly represented, but they definitely all contain Seuss' intellectual property.

Is it possible to put it on Github? If so, under what possible licenses?

1

tl; dr: Maybe, maybe not, but trouble seems likely.

Your text files are clearly derived from Dr. Seuss' books, so you normally need the permission of the copyright holder of the books.

You mention "fair use" as a possible way around this. However, I don't think this will necessarily fly. The releveant regulation says:

Notwithstanding the provisions of sections 106 and 106A, the fair use of a copyrighted work, including such use by reproduction in copies or phonorecords or by any other means specified by that section, for purposes such as criticism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research, is not an infringement of copyright. In determining whether the use made of a work in any particular case is a fair use the factors to be considered shall include—

(1) the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;

(2) the nature of the copyrighted work;

(3) the amount and substantiality 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.

[...]

17 U.S. Code § 107 - Limitations on exclusive rights: Fair use So the main factors are:

  • the kind of use intended
  • the type of work
  • the amount used
  • the effect upon the market value for the work

The kind of use (research) is probably not a problem, as is the type of work (a published book).

However, the amount used (several/all the books) would speak against "fair use". Using some small excerpts from the books may be okay, but using all of them probably not.

One mitigating factor: The effect upon the market could be argued to be small, since you will not publish the illustrations, and the Dr Seuss books are arguable quite incomplete without them.

In summary: You may get lucky with the fair use defense, but I would not count on it. Using the book texts for research yourself could be okay, but publishing it most likely is not.


Additional reading:

The article Text and Data Mining and Fair Use in the United States addresses your problem, and mentions some court cases. The fair use defense was upheld in some cases, but none of these published complete works, only excerpts.

  • So no-go on the plaintext IP. What about a slightly zanier branch of options, where I would encrypt the file contents (even a super simple Caesar cipher with a +/-1 shift would likely suffice). I could store the files encrypted on github, and just add a level of complexity for those who utilize my repo. Either they have to run a batch process to decrypt all the IP initially, or the entire RNN works with the encrypted data, and then its samples are decrypted into something legible. – Bango Mar 13 '17 at 16:55
  • Option 1 in comments above would mean that at some point, the users have access to the IP. Option 2 would mean that there is no IP ever being infringed on, as the machine is training on encrypted data, and its sample data is merely a mess of nonsense at face value, until it runs it through a simple decryption protocol. – Bango Mar 13 '17 at 17:00
  • @Bango: Sorry, no idea, though I personally don't think it would fly. That would make a good new question, though :-). – sleske Mar 13 '17 at 20:12
  • Answer: No. If it can't be shared unencrypted, it cannot be shared encrypted. link – Bango Mar 14 '17 at 17:43
  • @Bango: Yes, good find. – sleske Mar 14 '17 at 18:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.