Final1000 Dataset
This directory contains the Final1000 evaluation set used for NedoTurkishTokenizer.
Files
| File | Description |
|---|---|
nedo_final1000_gold.jsonl |
Main gold set. One JSON object per line. |
nedo_final1000_gold.csv |
CSV version of the same data. |
zenodo_metadata.json |
Draft metadata for a Zenodo release. |
Content
Final1000 is a 1,000-item evaluation set for Turkish morphological tokenization.
Source distribution:
| Source | Count |
|---|---|
annotator_agreed |
754 |
adjudicated_new |
196 |
adjudicated_old |
50 |
| Total | 1000 |
The build check reports 0 invalid examples.
Usage
import json
from pathlib import Path
path = Path("dataset/nedo_final1000_gold.jsonl")
examples = [json.loads(line) for line in path.read_text(encoding="utf-8").splitlines()]
print(len(examples))
Result files
Frozen metrics:
Error analysis:
Note
External links
| Resource | Link |
|---|---|
| GitHub repository | NMSOfficial/NedoTurkishTokenizer |
| Hugging Face dataset | Ethosoft/NedoTurkishTokenizer-Final1000 |
| Zenodo DOI | 10.5281/zenodo.21274980 |