--- dataset_info: features: - name: id dtype: int64 - name: text dtype: string - name: metadata dtype: string - name: region_main_id dtype: int64 - name: region_main dtype: string - name: region_sub_id dtype: int64 - name: region_sub dtype: string - name: date_str dtype: string - name: date_min dtype: float64 - name: date_max dtype: float64 - name: date_circa dtype: float64 - name: last_digit dtype: string - name: __index_level_0__ dtype: int64 splits: - name: train num_bytes: 52821262 num_examples: 70682 - name: validation num_bytes: 6688781 num_examples: 8751 - name: test num_bytes: 6514866 num_examples: 8769 download_size: 30227497 dataset_size: 66024909 configs: - config_name: default data_files: - split: train path: data/train-* - split: validation path: data/validation-* - split: test path: data/test-* license: apache-2.0 tags: - ancient languages size_categories: - 10K This is the I.PHI dataset from the Nature article *"[Restoring and attributing ancient texts using deep neural networks](https://www.nature.com/articles/s41586-022-04448-z)"*. I.PHI contains geographical and chronological metadata for a cleaned subset of the [Packard Humanities Institute](https://inscriptions.packhum.org/) database of ancient Greek inscriptions. ## Dataset Details ### Dataset Description - **Curated by:** Thea Sommerschield\*, Yannis Assael\*, Brendan Shillingford, Mahyar Bordbar, John Pavlopoulos, Marita Chatzipanagiotou, Ion Androutsopoulos, Jonathan Prag, Nando de Freitas - **Shared by:** George Baker - **Language(s) (NLP):** Ancient and Byzantine Greek - **License:** Apache 2.0 ### Dataset Sources - **Repository:** [github.com/sommerschield/iphi](https://github.com/sommerschield/iphi) - **Paper:** [Restoring and attributing ancient texts using deep neural networks](https://www.nature.com/articles/s41586-022-04448-z) ## Uses I.PHI contains data for chronological attribution (dating), geographical attribution, and text restoration of Ancient and Byzantine Greek inscriptions. #### Data Collection and Processing Duplicate inscriptions and inscriptions with fewer than 50 characters are dropped. Note: this version contains slightly more examples than the number reported in the original article (70,682/8,751/8,769 instead of 63,014/7,783/7,811). This is likely due to additions to the [Packard Humanities Institute](https://inscriptions.packhum.org/) database from which the dataset is sourced. ## Citation **BibTeX:** ``` @misc{sommerschield2021iphi, title={{I.PHI} dataset: ancient Greek inscriptions}, author={Sommerschield*, Thea and Assael*, Yannis and Shillingford, Brendan and Bordbar, Mahyar and Pavlopoulos, John and Chatzipanagiotou, Marita and Androutsopoulos, Ion and Prag, Jonathan and de Freitas, Nando}, howpublished = {\url{https://github.com/sommerschield/iphi}}, year={2021} } @article{assael2022restoring, title={Restoring and attributing ancient texts using deep neural networks}, author={Assael, Yannis and Sommerschield, Thea and Shillingford, Brendan and Bordbar, Mahyar and Pavlopoulos, John and Chatzipanagiotou, Marita and Androutsopoulos, Ion and Prag, Jonathan and de Freitas, Nando}, journal={Nature}, volume={603}, number={7900}, pages={280--283}, year={2022}, publisher={Nature Publishing Group UK London} } ``` ## Dataset Card Authors This dataset card was written by George Baker. ## Dataset Card Contact george (dot) baker (at) colorado (dot) edu