A Computational Expedition into the Undiscovered Country - Evaluating Neural Networks for the Identification of Hamlet Text Reuse
In this article, we describe a two-step processing pipeline for identifying text reuse of Shakespeare’s Hamlet in a corpus of postmodern fiction by comparing n-grams from both sources. A key feature of our approach lies in a pre-filtering step, in...
more
|
Full text:
|
|
|
Link for citation:
|
|
|
In this article, we describe a two-step processing pipeline for identifying text reuse of Shakespeare’s Hamlet in a corpus of postmodern fiction by comparing n-grams from both sources. A key feature of our approach lies in a pre-filtering step, in which we select target sentences in the fiction corpus that are potential candidates for Hamlet text reuse. Without pre-filtering, the amount of text reuse pairs (that are no actual quotes) would be extremely high. In a second filtering step, we compare potential text reuse pairs by their vector representation using a neural network trained in an unsupervised manner. We found that using the vector similarity produces a problematic amount of false positives. The created vector representations are created using an unsupervised training approach, resulting in similarity aspects that are unfavorable for our use case.
|
| Export to reference management software |
|
