In the previous episode of our new mini-series on computational creativity we have introduced a tale-generation algorithm which follows a logic/qualitative paradigm.
The computer is told what to do and how to do it, and it executes. In this way, the performance similarity to the human performance is directly proportional to the elegance and complexity of the written code (the ability of the developer).
In this episode the focus is on machine learning, an evolutionary/quantitative computational approach through which a huge bunch of data (in this case natural-language-written texts, the so-called corpus) , is processed. The generated texts show words in a probabilistic not logical way! Nonetheless the more data are taken into account, the more probability converges towards logics.
Short poetries result far more convincing compared to, for instance, presidential speeches, where the logic structure is a compulsory aspect to be considered.
Here follows a brief and incisive example of the poeticalness´ degree a machine is able to achieve.
If you have a deeper interest in the concepts underlying the automatic poetry generation you can get many useful insights by the following sources:
Selected and described by Cosimo Palma, Communication Trainee at the Terminology Coordination Unit of the European Parliament (Luxembourg).