|
--- |
|
license: cc0-1.0 |
|
language: |
|
- de |
|
tags: |
|
- natural-language-processing |
|
- poetry-generation |
|
- text-generation |
|
- torch |
|
- lstm |
|
--- |
|
|
|
This generative text model was trained using [Andrej Karpathy's code](https://github.com/karpathy/char-rnn) on texts of German [Hexameter](https://en.wikipedia.org/w/index.php?title=Hexameter&oldid=1198526559). Models of this type well represent individual style. |
|
|
|
Model was trained with size 512 and 3 layers, dropout 0.5. |
|
|
|
## Usage |
|
|
|
The procedure for installing the required software is described [by Karpathy](https://github.com/karpathy/char-rnn), torch is required, the code is written in lua. Be careful, versions of libraries written many years ago are used! |
|
|
|
```bash |
|
th sample.lua hexameter_lm_lstm_epoch80.00_1.3702.t7 |
|
``` |
|
|
|
## Train data |
|
|
|
Hexameter lines extracted from [large collection of German verses](https://github.com/tnhaider/DLK) running by Thomas Haider. |
|
|
|
## What for? |
|
|
|
In an era of winning Transformers, ancient RNN models seem archaic. But I see that they still work better than modern architectures with such important categories from the humanities point of view as poetic style. |
|
|
|
## Dataset |
|
|
|
There is a [repository](https://github.com/nevmenandr/german-generated-poetic-texts) that publishes German poetic texts generated by RNN LSTM models (inclding this one) with different temperature. |
|
|
|
## Publication |
|
|
|
There are some texts explaining the goal o these poetic experiments and their place in the history of human culture. |
|
|
|
* "[Der digitale Superdichter](http://nevmenandr.net/personalia/holderlin.pdf). Vor 250 Jahren wurde Friedrich Hölderlin geboren. Heute kann Computertechnik neue Gedichte im Hölderlin-Sound generieren. Ein Werkstattbericht" *Die Literarische Welt*, 14 March 2020, p. 29. (included in this repository as pdf) |
|
|
|
* Orekhov, Boris, and Frank Fischer. "Neural reading: Insights from the analysis of poetry generated by artificial neural networks." *Orbis Litterarum* 75.5 (2020): 230-246. [DOI: 10.1111/oli.12274](https://onlinelibrary.wiley.com/doi/abs/10.1111/oli.12274) |
|
|
|
## BibTeX entry and citation info |
|
|
|
``` |
|
@article{orekhov2020neural, |
|
title={Neural reading: Insights from the analysis of poetry generated by artificial neural networks}, |
|
author={Orekhov, Boris and Fischer, Frank}, |
|
journal={Orbis Litterarum}, |
|
volume={75}, |
|
number={5}, |
|
pages={230--246}, |
|
year={2020}, |
|
publisher={Wiley Online Library} |
|
} |
|
``` |
|
|
