dknews-GPT3
所属分类：GPT/ChatGPT
开发工具：Jupyter Notebook
文件大小：139KB
下载次数：0
上传日期：2023-01-02 21:34:27
上 传 者：
sh-1993
说明： [@MinaAlmasi]（https:github.com MinaAlmosi）和[@drasbaek]（https:github.com drasbaek）关于GPT-3和丹麦语合成新闻生成的项目
(project on GPT-3 and synthetic news generation in Danish by
[@MinaAlmasi](https://github.com/MinaAlmasi) &
[@drasbaek](https://github.com/drasbaek)
,)
文件列表:
code (0, 2023-01-03)
code\gpt-3 (0, 2023-01-03)
code\gpt-3\article-generation.ipynb (16239, 2023-01-03)
code\gpt-3\fine-tuning.ipynb (11396, 2023-01-03)
code\gpt-3\finetune-results-gpt.csv (100189, 2023-01-03)
code\gpt-3\training-loss-gpt.Rmd (2793, 2023-01-03)
code\machine-classifiers (0, 2023-01-03)
code\machine-classifiers\bert-train-test.ipynb (133978, 2023-01-03)
code\machine-classifiers\bow-train-test.ipynb (41142, 2023-01-03)
code\machine-classifiers\tfidf-train-test.ipynb (41404, 2023-01-03)
code\real-article-processing (0, 2023-01-03)
code\real-article-processing\danewsroom-article-process-and-shorten.ipynb (23773, 2023-01-03)
code\real-article-processing\scraped-article-preprocess.Rmd (1741, 2023-01-03)
code\real-article-processing\scraped-article-shorten.ipynb (8789, 2023-01-03)
code\results (0, 2023-01-03)
code\results\classifier-agreement.Rmd (871, 2023-01-03)
code\results\experiment-A (0, 2023-01-03)
code\results\experiment-A\experiment-A-inspection.Rmd (20342, 2023-01-03)
code\results\experiment-B (0, 2023-01-03)
code\results\experiment-B\bert-finetune-table.Rmd (1484, 2023-01-03)
code\results\experiment-B\classifier-comparison-table.Rmd (3465, 2023-01-03)
test-data (0, 2023-01-03)
test-data\all-classifier-predictions.csv (78585, 2023-01-03)
# Cognitive Science Bachelor's Project on GPT3 & Danish News Articles
@MinaAlmasi & @drasbaek
### Content overview
- **code** (contains all code files from this investigation both for fine-tuning GPT-3, building classifiers and analyzing experimental results)
- **test-data** (contains file with the 96 test articles from both experiments and how they were classified)
### Dataset of scraped news articles
- For copyright reasons, the dataset for training the machine classifiers is not publicly available on this GitHub repository. For more information about this, please write mina.almasi@post.au.dk or 202008161@post.au.dk
### Fine-Tuned BERT Model
Our trained classifier for detecting synthetic news is available on the *Hugging Face Hub* on this link:
* https://huggingface.co/MinaAlmasi/dknews-NB-BERT-AI-classifier
The BERT model is a fine-tuned version of [NB-BERT-LARGE](https://huggingface.co/NbAiLab/nb-bert-large). More details regarding our model is available on the aforementioned link.
近期下载者：
相关文件：
收藏者：