README.md

---
license: mit
language:
- la
base_model:
- latincy/la_core_web_lg
model-index:
  - name: la_core_web_lg_3.7.4
    results:
      - task:
          type: NER
        dataset:
          name: Herodotos_dataset
          type: Herodotos_dataset
        metrics:
          - name: macro F1
            type: macro F1
            value: 58
        source:
          name: SEFLAG
          url: https://bibbase.org/network/publication/schulz-deichsler-seflagsystematicevaluationframeworkfornlpmodelsanddatasetsinlatinandancientgreek-2024
      - task:
          type: lemmatization
        dataset:
          name: UD-Latin
          type: UD-Latin
        metrics:
          - name: accuracy
            type: accuracy
            value: 88
        source:
          name: SEFLAG
          url: https://bibbase.org/network/publication/schulz-deichsler-seflagsystematicevaluationframeworkfornlpmodelsanddatasetsinlatinandancientgreek-2024
---
**la_core_web_lg**

- **Person or organization developing model**: [Patrick J. Burns; with
Nora Bernhardt \[ner\], Tim Geelhaar \[tagger, morphologizer, parser,
ner\], Vincent Koch \[ner\]](https://diyclassics.github.io/)

- **Model date**: May 2023

- **Model version: 3.7.4**

- **Model type:** spaCy

- **Information about training algorithms, parameters, fairness
constraints or other applied approaches, and features:** For information on the training workflow see p.4-5 of LatinCy: Synthetic Trained Pipelines for Latin NLP
(https://arxiv.org/pdf/2305.04365v1)

- **Paper or other resource for more information:** *Burns, P.J. 2023.
"LatinCy: Synthetic Trained Pipelines for Latin NLP." arXiv:2305.04365
\[cs.CL\]. http://arxiv.org/abs/2305.04365.*

- **License:** *MIT*

- **Where to send questions or comments about the model:**
https://diyclassics.github.io/

Intended Use

- Primary intended uses: Morphological analysis, POS-Tagging,
Lemmatizing, Parsing, NER

- Primary intended users: Classical Scholars

- Out-of-scope use cases: unknown

Data, Limitations, and Recommendations

- Data selection for training: Training data consists of latin
UD-Treebanks, Wikipedia and OSCAR sentence data, the CC-100 Latin
dataset and the Herodotos Project NER dataset

- Data selection for evaluation: Evaluation was done according to the
spaCy workflow and is documented in the meta.json file found in the
repository
(https://huggingface.co./latincy/la_core_web_lg/blob/main/meta.json)

- Limitations: unknown